xv De. Alan A. Desrochers of Rensselaer Polytechnic Institue, Troy, New York; Dr. Omar Farooq of Aligarh Muslim University, India; Dr. Rajendra S. Gad of Goa University, India; Dr. Vikram M. Gadre of India Insitute of Teckoology, Bombay; Dr E.Gopinathan of National Insitute of Technology. Calicut, India Dr. Lim Heng-Siong of Multimedia University, Malayasa; Dr, Hamid Gholam Hosseini of Auckland Uni versity of Technology, New Zealand; Dr. J. Abdul Jaec! of Thangal Kunju Musaliar College of Engineer: ing. Kerala, India: De Sabira Khatun of University Purra Malaysia; Dr.Ju Liu of Shandong University (China: De Wagdy H. Mahmoud of University of the District of Columbiz; Dr. Ashutosh Marathe of Vist. wakarma Insitute of Technology, India; Dr S\N, Merchant of Indian Institute of Technology, Bombay Dr, Muhammad Javed Mirza of Riphah International University, Pakistan: Dr. Ravinder Nath of National {nsttue of Technology, Hamirpur, India; Dt K MM. Prabhu of Indian institute of Technology. Madras Dr. SM. Sameer of National tnstitue of Technology, Calicut, India; Dr.P Sahu of National Institute of Technology, Kurukshetra, India; Dr. Mansour Tabernezhadi of Northem Illinois University: Dr. Nisachon Tangsangiumvisai of Chulalongkorn Univesity, Thailand: and Dr. imran A. Tasaddug of FAST:National University of Computer & Emerging Sciences, Karachi, Pakistan, The manuscript ofthe fourth edition was reviewed by Dr. Wolfgang FG. Mecklenbrauker of Technical University of Vienna, Austria, nd Dr. Truong Nguyen of University of California, San Diego. Some parts ofthe manuscript of the fourth edition were reviewed by Dr.Subash Dutta Roy of Indian Lastitute of Technology, New Delhi {thank all of them for their valuable comments, which have improved the book tremendously a mante amet research tudes reviewed vations portions ofthe manuscript of al editions and Rafer Gans cat MATA porn: In parca, woud ite o tank Drs. Chares D.Creuser, Sone Lins Les Lane, Sean Hapoel, Zhi He, Hsn-Han Ho, Michal Lightstone, In8- and Mylene Quite de Fara a ose baeME Maker, Norbert Strobe, Stefan Thurmhofer CE 138d ECE 2A hn ME Be Leuk: Lam abo indebted to all omer std a ty class EE 483 at the University of Southern Cy seudonts ‘hich helped refine the book. Supplements All MATLAB programs incladed is from the Intemet site wwwece sn eine isbedu/Faculty A solutions manual prepared by te #08 to all problems and MAT; ‘Most materials ofthis ‘CD accompanying this book a itraBookse Hsin-Han Ho, Travis LAB exercises is avaiab book are available to isin are also available Smith, and Martin Gawecki and containing the eo instructors from the publisher, PowerPoint RctOrs from the author.Table of Contents Preface ix Acknowledgements 1 Signals and Signal Processing 1 1.1 Characterization and Classification of Signals 1 1.2. Typical Signal Processing Operations 4 13. ExamplesofTypical Signals. 13 14 ‘Typical Signal Processing Applications 21 15 Why Digital Signal Processing? 37 Discrete-Time Signals in the Time Domain 4 2.1 Time-Domain Representation 42 22 Operationson Sequences 46 23. Operations cn Finite-Length Sequences $5 24 ‘Typical Sequences and Sequence Representation 62 25 TheSampling Process 72 26 Correlation of Signals 74 27 Random Signals 80 28 Summary $1 29 Problems 81 2.10 MATLAB Exercises. 87 Discrete-Time Signals in the Frequency Domain 8 3.1 The Continuous-Time Fourier Transform 89 3.3 Discrete-Time Fourier Transform Theorems 105 34 Energy Density Spectrum of a Discrete-Time Sequence IL 35 Band-Limited Discrete-Time Signals 112 36 DTFT Computation Using MATLAB 113 3.7 The Unwrapped Phase Function 113 3.8 Digital Processing of Continuous-Time Signals. us ae39 Samplingof Bandpass Signals 129 3.10 Effestof Sample-and-Hold Operation 131 3.11 Summary 132 3.12 Problems 133 3.13 MATLABExercises 142 4 Discrete-Time Systems 143 4.1. Discrete-Time System Examples 143 42. Classification of Discrete-Time Systems 149 43° lmpute and Step Responses 153 Time-Domain Characterization of LTI Discrete-Time Systems 1 Simple Interconnection Schemes 161 $10 su siz saa sas sus ae EEE Finite-Dimensional LTI Discrete-Time Systems 164 (Casification of 11 Discrete-Time §; stems Frequency-Domain Representations of LT] Discrete-Time Systems Phase and Group Delays 185 Summary 189 Problems 199 MATLAB Exercises 198 rngth Discrete Transforms 199 99 The Discrete Fourier Transform 29] Kelanon Between he DTFT and the DET and Theis lnverses Circular Convolution 244 Classifications of Finite-Length Sequences 216 DET Symmetry Relations 3) Discrete Fourier Tran orm Theorems Fourier-Domain Filtcring 230 Computation ofthe DFT of Real Sequences Linear Copvolition Using the DET Short-Time Fourier Transform 234 24s Discrete Cosine Transform yy The Haar Transform 56 Escrty Compaction Froperies 959 Semmary 39 Problems 3g5.17 MATLAB Exercises 275 Transform 277 6.1 Definition 62 Rational =-Transforms 281 6.3 Region of Convergence of a Rational z-Transform 283 64 The Inverse z-Transform 289 6.5 s-Transform Theorems 7 6.6 Computation of the Convolution Sum of Finite-L 308 6.7 The Transfer Function 308 68 Summary 320 69 Problems 320 6.10 MarLas Exercises 332 7 LT Diserete-Time Systems in the Transform Domain 7.1 ‘Transfer Function Classification Based on Magnitude Characteristics 333 7.2 Transfer Function Classification Based on Phase Characteristics 342 7.3: ‘Types of Lincar-Phase FIR Transfer Functions 349 7A Simple Digital Filters ) 7.5 Complementary Transfer Functions 379 716 Inverse Systems 385 7.7 System Identification 389. 78 Digital TwoPairs 392 7.9 Algebraic Stability Test 394 7.10 Summary 399 7.11 Problems 400 7.12 MATLAB Exercises 44 Digital Filter Structures 417 8.1 Block Diagram Representation 418 82 Equiva 83 Basic FIR Digi Structures 421 Filter Structures 42: 84 Basic HR Digital Filter Structures 4 85 Realization of Basic Structures Using MATLAB 433, 86 Alipass Filters 436 8.7 Parametrically Tunable Low-Order IIR Digital Filter Pairs 445, 828 IIR Tapped Cascaded Lattice Structures 447 xvii89 FIRCascaded Latice Structures 452 8.10 Parallel Allpass Realization of IR Transfer Functions 460) 8.11 Tunable High-Order Digital Fits 465 \ 8.12 Computational Complenty of Digital Fier Structures 472 813 Summary 472 8.14 Problems 473, 8.15 MATLAB Exercives 487 9 TIR Digital Fiter Design 499 9.1 Prciminary Considerations 489 92 Bilinear Transformation Metiod of IR Filer Design 494 93 Design of Lowpass IR Digital Fiters 499 94 Desig of Highpas, Bandpass, and Bandstop IIR Digital Filters. SOL 9S Special Transformations of1IR Fess 503 6 MR Digital Fier Design Using MATLAB 512 97 Computer-Aided Design of IR Digital ters §15 98 Summary $19 99 Problems 519 9.10 MATLAB Exercises 595 10 FIR Digital Fiter Design 37 104 Preliminary Considerations $97 102 FIR Fite Design Based on Windowed Fourie Serie 331 103 Compate * Aided Desin of Equipe Linea: Phase FIR Filters of Minimam-Phase FIR Fite 105 FIR Digital iter Design 106 Desig 546 104 Design 585 Using Matias 556, of Computationally Efi 107 Summary $6 108 Problems 109 Mar FIR Digital Fiters 573 307 AB Exercises 554 DSP Algorithm Implementation 599 N11 Basic Issues 55g 112 Structure Simulation nd Verification Ui 3 : ‘ing MATLAB, Compa 610 on ofthe Discrete Fourier T ; ransform 647 114 Fast DET Algoritms Based on lode Mapping 63 S DFT. ‘oi }OFT Compustin Using Mating 0116 Sliding Discrete Fourier Transform 642 ns 19 11.10 wu M2 12 Analysis of 12.10 12.11 12.12 12.3 12.14 12.5 DFT Computation over a Narrow Frequency Band 642 Number Representation 647 Handlit Summary 653 of Overflow 652 Problems 653 MATLAB Exercises 661 nite Wordlength Effects 663 ‘The Quantization Process and Errors 664 Quantization of Fixed-Point Numbers 665 Point Numbers 668 lysis of Coefficient Quantization Effects 668 Quantization of Floatiny An A/D Conversion Noise Analysis. 681 Analysis of Arithmetic Round-Off Errors 691 Dynamic Range Scaling 695 Signal-to-Noise Ratio in Low-Order IIR Filters 706 Low-Sensitivity Digital Fiters 710 Reduction of Product Round-Off Noise Using Error Feedback Limit Cycles in IR Digital Filters 719 Round-Off Errors in FFT Algorithms 727 Summary 730 Problems 731 MATLAB Exercises 736 13 Multirate Digital Signal Processing Fundamentals 739 13. 3 B4 BS 136 137 Bs Bg 1340 ‘The Basic Sampling Rate Alteration Devices 740 Multirate Structures for Sampling Rate Conversion 750 758 Multistage Design of Decimator and Interpo ‘The Polyphase Decomposition 760 Arbitrary-Rate Sampling Rate Converter 771 Nyquist Filters 783 CAC Decimators and In ‘Summary 796 Problems 797 MATLAB Exercises 805 polators 792 116 xix14 Multirate Filter Banks and Wavelets 807 14.1 Digital Filter Banks 807 142 Two Channel Quadature-Mirror Fi Bank 813 143 Perfect Reconstruction Two-Channel FIR Filter Banks 144 L-ChannelQMF Banks 832 145 Mulilevel Filer Banks 840 46 Disereie Wavelet Transform 844 47 Summary 853 48 Problems 853 4.9 MATLAB Exercises $61 A Analog Lowpass Filter Design 863, A.| Analog Filer Specifications $63, A2 Butierworth Approximation 865 A3 Chebyshey Approximation 867 AA Elliptic Approximation 870 A. Linear Phase Approximation 871 A Analog Filter Design Using MATLAB 872 A.7 Analog Lowpass Filter Design Examples AS A Comparison of the Filter Types $77 A9 Anti-Aliasing Fier De 880 A.10 Reconstruction Filter Design 882 875 Design of Analog Highpas, BL Analog Hi B2 Anak Bandpass,and Bandstop Filters. $87 hpass Filter Design 7 Bandpass Filter Design 889 B3 Analog Bandstop Filter Design 92 © Discrete-Time Random Signals 93 C:1 Statistical Properties of a Random Variable erties of a Random Signal ‘Wide-Sense Stationary Random Signal 496 mceptof PowerinaRandom Signal 497 S Ergodic Signal 893 895 C2 Statistical Ps ae ® Transform-Domain Represenations of Random $i 899 7 White Noise 9g C8 Discrete-Time Pr Bibliography 997 Index 927 ing of Random Signals 901Chapter 1 Signals and Signal Processing Signals play an important role in our daily lives. Examples of signals that we encounter frequently are speech, music, picture and video signals. A signal isa function of independent variables such as time. dis- tance, position, temperature, and pressure. For example, speech and ignals represent air pressure as a function of time at a point in space. A black-and-white picture is a representation of light intensity as a function of two spatial coordinates, The video signal in television consists of a sequence of images. called frames, and is a function of three variables: two Most signals we encounter are generated by natural means. However, a signal can also be generated synthetically or by computer simulation. A signal carries information, and the objective of signal provess- ing is to extract useful info ied by the signal, The method of information extraction depends on the type of signal and the nature of the information being carried by the signal, Thus, roughly speak: ing, signal processing is concerned with the mathematical representation of the signal and the algorithmic ‘operation carried out on it to extract the information present. The representation of the signal can be in terms of basis functions in the domain of the original independent variable(s), or itcan be in terms of basis functions in a transformed domain. Likewise, the information extraction process may be carried out in the original domain of the signal or in a transformed domain. This book is concerned with discrete-time representation of signals and their discrete-time processing This chapter provides an overview of signals and signal processing methods, The mathematical char ‘acterization of the signal is first discussed along with a classification of signals, Next, some typical signals ‘are discussed in detail, and the type of information cartied by them is described. Then, a review of some ‘commonly used signal processing operations is provided and illustrated through examples. A brief review of some typical signal processing applications is discussed next, Finally, the advantages and disadvantages of digital processing of signals are discussed. ial Coordinates and time. ation c 1.1 Characterization and Classification of Signals Depending on the nature of the independent variables and the value of the function defining the signal various types of signals can be defined. For exampie. independent variables can be continuous or dis crete, Likewise, the signal can either be a continuous or a discrete function of the independent variables. Moreover, the signal can be either a real-valued function or a complex-valued function. A signal can be generated by a single source or by multiple sources. In the former case, itis a sealar signal, and in the latter case, itis a vector signal, often called a multichannel signal. A one-dimensional (1-D) signal is a function of a single independent variable, A two-dimensional (2-D) signal is a function of ‘wo independent variables, A multidimensional (M-D) signal is @ function of more than one variable. The speech signal is an example of a 1-D signal where the independent variable is time. An image signal, suchSpeech Demo 1 Image Demo 1 Video Demo 1 Chapter 1; Signals and Signal Processing xpos sm cxmape of 2D sian wer the wo independent warble ar the two spatial ee i see 2 ieee pel hte 5 fction of v0 vel ati el oe fm cng vont roi of ine. Hence, ‘black-and-white video signal can be considered an example ofa three-dimensional (3-D) signal whe thee independent variables are the two spatial variables and time. A color video signal composed of three 3-D signals representing the three primary colors: red, For transmission purposes, the RGB television signal is transforme signal composed ofa luminance component and two chrominance components. ie. ls The value of the signal at a specific value of the independent variable is called its amplitude. The ‘variation of the amplitude as a function ofthe independent variable is called its waveform. ; For 0 1-D signal, the independent variable is usually labeled as rime. If the independent variable is continuous, the signal is called a continuous-time signal. If the independent variable is discrete, the signal is called a discrete-rime signal. A continuous-time signal is defined st every instant of time. On the other hand, a discrete-time signal takes certain numerical values at specified discrete instants of time and between theve specified instants of time, the signal is not defined. Hence, a discrete-time signal is basically a sequence of numbers A continuous-time sign 1 is a three-channel een, and blue (RGB) ther type of three-channel into i with a continvous amplitude is usually called an analog signal. A speech of an analog signal. Analog signals are commonly encountered in our daily lives and are usually generated by natural means. A. isrete-time signal with dscrete-valued amplitudes rep serced by a fine number of digits is referred oa a digital signal. An example ofa digital signal is the tietiesd music signal stored in a CD-ROM disk. discrete-time signal with continuous-valved amp {oies i called a sampled-data signal This last type of signal occurs in switched capacitor (SC) cireuits. ized sampled-data signal. Finally, a continuous-time signal with discrete red 10 a8 a quantized boxcar signal ($ie93). The latter type of signals curs in digital electronic circuits where the signal is kept at fixed level (usually one of two values) between two instants of clocking. Figure 1.1 ih ustrates the four types of signals, ‘The functional dependence of signal in ts mathematical representation is often a daca astne ED signal, the continuous independent variable is usually denot 8 discrete-time 1-D signal, the discrete independent variable i uscal al and {vn} epresents a continuous-time 1-D si plicitly shown. For ied by 1, whereas for y denoted by n, For example, u(t) represents a discrete-time 1-D signal, Each member, ple. I many applications, a discret. ated -time signal is gene y sampling the later at uniform intervals of time, Ifthe discrete fined are uniformly spaced, the independent discrete to assume integer values sch tinwous- time 2-D signa, the two independent variables are usual ig ° nt variables are usually the spatial Can be oop wally dence by x and y. For example the inten ofa tee and- white imag Primary cena MCE?) A color image u(r). 1 compared of thre signe ing the Primary colors, red, green, and blue 7 fist ia rx,9) u(r») =| gix.y) d(x, y) fn], of a discrete-time signal is called from a parent coatinvous- On the other hand, ad discretized spatial variables lon). Likewise, a black-and where x and y denote the two D discrete signal, and its two independent Variable pc and its two independent variables arc Lahhecd by m and n. Hence, digitized image tan be represented as spatial ero oauesce isa 3-D signal and can be sepresenscn as u(x, y,t), *paial variables and + denotes the temporal vatiabhe tine A color video1.1. Characterization and Classification of Sign \ ampl > Time, f (@) ) Ne i =| a © @ Figure 1.1; (a) An analog signal, ) a digital signal, (c) a sampled-data signal, and (4) a quantized boxcar signal signal is a vector signal composed of three video signals representing the three primary colors, red, green and blue. ‘There is another classification of signals that depends on the certainty by which the signal can be uniquely described. A signal that can be uniquely determined by a well-defined process such as a math- ematical expression or rule, oF table look-up, is called a deterministic signal. A signal that is generated in.a random fashion and cannot be predicted ahead of time is called a random signal. In this text, we are primarily concemed with the processing of discrete-time deterministic signals. However, since practical discrete-time systems employ finite wordlengths for the storing of signals and the implementation of the signal processing algorithms, it is necessary to develop tools for the analysis of finite wordlength effects ‘on the performance of discrete-time systems. To this end, it has been found convenientto represent certain pertinent signals as random signals and employ statistical techniques for their analysis. ‘Some typical signal processing operations performed on analog signals are reviewed in the following4 Chapter 1: Signals and Signal Processing 1.2 Typical Signal Processing Operations in and frequency-domain operations are employed. In cither case, the desired ‘operations are implemented by a combination of some elementary oper ions. These operations are also ly implemented in real-time or near real-time, even though, in certain applications, they implemented off-line ay be 1.2.1. Simple Time-Domain Operations The three most basic time-domain signal operations are scaling, delay, and addition. Scaling is simply the ‘uliplication ofthe signal bya positive or a negative constant. In the case of an is usally called amplification if th cone. If the magnitude of the multi Ths, if x(¢) i an ignals, this operation nitude of the multiplying constant, called gain, is greater than ng constant is less than one, the operation is called attenuation. ialog signal. the scaling operation generates a signal y(t) = x(t), where a is the ‘multiplying constant The delay operation generates a signal that is a delayed re Signal x(t), y(0) = x(t ~to) i the signal obtained by del Vo be a postive number. If is negative, the Many applications require oper Sea ag i) +2200 — an) isthe signal pened by the addition ofthe three analog signals two ipa ey: Amor elementary operation isthe product of two signals. Thus, the product ‘wo signals x(t) and x2(¢) generates a signal y(t) = x(t) 19(0) Two other elementary operations are integration and differestiation, x(t) generates a signal y(i) =f de(y/dt plica of the original signal, For an analog laying x(t) by the amount ‘9, which is assumed nit is an advance operation. rations involving two or more signals to generate a new signal. For The integration of an analog s(t) dr, while its differentiation results i signal w(t) = The first three clementary ope: discn da rations, namely, scaling, delay, and addition, discussed Further in later pe i differentiation are im view some commonly used! complex F more ofthe elem better understood in the frequer c-time signals a also carried out on ‘Sof this text, The other two elementary opera. plemented approximately in the discrete-time domain. al processing operations that are implemented by NAEY Operations. The characteristics of some of these operations are 3 domain by making use of the continuows-lime Fourier treesform, The transform X(j2) time signal x(t) is given by! tions, namely, integrator Next we re ‘combining two ) of «continuous xg XQ) is ca led the spectrum of x(t) 12.2 Filtering ‘One ofthe most widely sed complen si Wo ales the spec ignal processing operations is filtering, er, Far cea lit Soe given specifications. The system : For example, the filter may be designed to the system and 10 block other frequer called a through whose main objective is n implementing this operation is requency components in a signal ange of frequencies of the signal 8S certain fn ¥ components. The1.2, Typical Signal Processing Operations 5 components allowed to pass through the filter is called the ‘components blocked by the filter is called the stopband. Various types of filters can be defined, assband, and the of frequencies of the depending on the nature of the filtering operation, In most cases, the filtering operation for an als is performed by a linear, time-invariant filter. If the filter is characterized by an impulse response /i(?) then its output y(¢) to an input is given by the convolution " ve) = [Meats 2) assuming the filter is relaxed with zero initial conditions at the time of application of the input signal, In the frequency domain, the above equation can be expressed as ¥(JQ) = HUQ)XUD), a3) where ¥(j2), X(j2), and H(j22), are, respectively, the continuous-time Fourier transforms of y(t), x(0),and h(t) A Jowpass filter passes all low-frequency components below a certain specified frequency fp. called the passhand edge frequency, and blocks all high-frequency components above f,, called the siopband edge frequency. A highpass filter passes all high-frequency components above a certain passband edge frequency f, and blocks all low-frequency components below the stopband edge frequency fy. A band: ‘pass filter passes all frequency components between two passband ed eS fp and Syn where fri < fa» and blocks all frequency components below the stopband edge frequency f,, and above the stopband edge frequency Jr2. A bundsiop filter blocks all frequency components between two stopband edge frequencies fy, and fz and passes all frequency components below the passband edge frequency fyx and above the passband edge frequency fy2. Figure 1.2¢a) shows a signal composed of three sinu: soidal components of frequencies 50 Hz, 100 Hz, and 200 Hz, respectively. Figures 1 2(b) to (e) show the results of the above four types of filtering operations with appropriately chosen cutoff frequencies. ‘A bandstop filter designed 0 block a single frequency component is called a notch filter. A mult ‘and filter bas more than one passband and more than one stopband. A cont filter is designed to block frequencies that are integral multiples of a low frequency ‘A signal may get corrupted unintentionally by an interfering signal called interference or noise, In ‘many applications, the desired signal occupies a low-frequency band from dc to some frequency fi, He, and itis corrupted by a high-frequency noise with frequency components above fv Hz with fir > fi In such eases, the desired signal can be recovered from the noise-cormupted signal by passing the latter through a lowpass filter with a cutoff frequency f, where fi, < fe < fy. In some applications, the noise corrupting the desired signal may be a single-frequency sinusoidal signal. For example, the noise generated by power lines radiating electric and magnetic fields appears as a 60-H.z sinusoidal signal. The desired signal can then be recovered by passing the corrupted signal through 4 noteh filter with 2 notch frequency at 60 Hz? 1.2.3 Generation of Complex-Valued Signals As indicated e called a real signal, while the latter is called a compler signal. AM naturally generated signals are ied, In some applications, it is necessary to develop a complex signal from a real signal having more ier.a signal can be real-valued or complex-valued. For convenience, the former is us a many countries, powerlines generate 5-H meChapter 1 Signals and Signal Processing 0 % «0 6 a 10 Time, msec TTT TTA “ll nM Wy HAL 2 40 6) 100 Time, msec © ©) ter ouput Blandstop fier output TI 2 TI Vth] 2 an 20406080100 Time, mee a Pree 12: fpr _— He, (d) output of a bandpass iter with cutofis at 80 Hz a Pare poeven 4 150 Hz, and ¢e wtpet of a bandstop filter ‘desirable properties. One ‘ es. One approach t Hilbert transformer t if & Complex signal from a aby "Scharacterized by an impulse response hyn Teal signal is by employing a en by [Fre94}, [Opps3} 1 fuee(t) =1.2. Typical Signal Processing Operations 7 Figure 1.3: Generation of an analytic signal using a Hilbert transformer. with a continuous-time F rier transform Hyrr(j82) given by =f. 250, Hyr(j2) = Dak as j < Let.x(¢) denote areal analog signal with a continuous-time Fourier transform X(/22). The magnitude spectrum of areal signal exhibits even symmetry, while the phase spectrum exhibits odd symmetry. Thus, the spectrum X(/@) ofa eal signal x(t) contains both positive and negative frequencies and can therefore be expressed as X(jQ) = Xp(JD) + Xalj2). a6 where Xp(j£2) isthe portion of X(j2) oceupying the positive frequency range and X, (22) is the portion of X(/2) occupying the negative frequency range. If x(¢) is passed through a Hilbert transformer, its ‘output £(1) has a spectrum X(/32) given by U2) = Myr JQ)XUA) = ~]X (4) + jXnl2) a7 It can be shown that £(F) is also a real signal, Consider the complex signal y(t) formed by the sum of x(0) and £(0): yf) =x) + J800), aay ‘The signals x(t) and £(¢) are called, respectively, the in-phase and quadrature components of y(t). The continuous-time Fourier transform of y(t) is given by (Problem 3.9) YUM) = XUQ) + JX U2) = 2X,A). a9) Thus, the complex signal y() called an anaiyric signal, has only positive-frequency components ‘A block diagram representation of the scheme for the analytic signal generation from a real signal is sketched in Figure 1.3, One application of the Hilbert transformer is in the implementation of a single- sideband modulation as indicated in Figure 1.8 and discussed in Section 1.2.4 1.2.4 Amplitude Modulation For transmission of signals over long distances, a transmission medium such as cable, optical fiber, or the atmosphere is employed. Each such medium has a bandwidth that is more suitable for the efficient transmission of signals in the high-frequency range. As a result for the transmission of a low-frequency signal over a channel, it is necessary to transform the signal to a high-frequency signal by means of a modulation operation. At the receiving end, the modulated high-frequency signal is demodulated, and the desired low-frequency signal is then extracted by further processing. There are four major types of modulation of analog signals: amplitude modulation, frequency modulation, phase modulation, and pulseChapter 1; Signals and Signal Processing 2) 2-2, w ty the modulated signal y(t). For Figure 14: (a) Spectrum of the modeling signal x(t) and (b) spectrum of the modulated smvenience, both spectra ae shown as rel fubctions amplitude modulation. Of these schemes, amplitude modulation is conceptually simple and is discussed r [Fre], [Opp83} “i Inthe ample modulation scheme, the amplitude ofa high-frequency sinusoidal signal A cos(2st), called the carrier signal, i varied by the low frequency band-limited signal x(1), called the modulating signal, generating a high-frequency signal, called the mo ated signal, y(t) according to vit) = Ax) c0s(250), (10) Thus, amplitude modulation can be iny the carrer signal. It can be shown plemented by forming the product ofthe modulating signal with that the spectrum ¥(/22) of y(t is given by YUQ) = $X (0-25) + 4x (112 + 9,)) aay where X(/2) isthe spectrum of the moc tlating signal and the modulated signal under the ass than 2, the highest frequency contained in x() high-frequency signal with a bandwidth The portion ofthe amplitude-modula Whereas the portion between 2, and £2, of two sideband ing signal x(t). Figure 1.4 shows the spectra of the mod. iption that the carrier frequency 2, is greater As seen from this figure, y(t) is now a 2m centered at 2p ted signal between 9, 2 and the absence of a carrier ouble-sideband supp \d-limited and $2y-+ 2y_ is called the upper sideband, im is called the lower sideband. Because of the ‘component in the pressed carrier (DSB-SC) modulation, he demodulation of Y(t) assuming 2, > Qy.is cared out in two s ‘ith sinascidal signal ofthe same frequency as the yeneration al, the process is called First, the pr Cartier is formed. This results in = Ax(t) cos? yt ne rewritten as (C05 Dot = 4x40) + 4x66) cos( ) a3) This result indicat factor 1/2 and an an i as indicated that the product signal is comy tude-modulated signal with Figure | S. The original modula through a lowpass filter with a cutoft fn ‘output of the posed of the origi carter frequency 2 signal can now be: tequency 2 satisfying the re 4 replica ofthe modulating signal iagram representations of assumption i I modulating si ignal scaled by a o- The spectrum R(j2) of r(t) fecovered from r(1) by passing it lation 2m < 2e < 22, ~ 2m. The er is then a sca Figure 1.6 shows the block schemes. The under f the amplitude mo the demodulation proc: odulation and demodul 288 outlined above is that a sinusoidal1.2. ‘Typical Signal Processing Operations 9 a a, | 2, ° 2 (22-2,,) 22-2, Figure 1.8: Spectrum ofthe pr oat oO Mt hot duct of the modulated signal and the carrier seos cos 0, @ © Figure 1.4: Schematic represeattions ofthe DSB-SC smpltnde modulation and demodulation schemes (a) mesltr and b) demodulator Maalting signal Modulted cai ~ / 4 / sy 4 | f i A tl geese Navy hate a I mn i ih E Ul / a \ 2 Seo spare soe Nhs in at ee a a a ee Time, mee Tankee @ b) Figure 1.7: (a) A sinusoidal modulating signal of frequency 20 Hz and (b) modulated carrer with a carrier frequency of 400 Hz based on the DSB modulation signal identical to the cartier signal can be generated at the receiving end. In general, it is difficult to ensure that the demodulating sinusoidal signal has a frequency identical to that of the carrier all the time. ‘To get around this problem, in the transmission of amplitude-modulated radio signals, the modulation process is modified so that the transmitted signal includes the carrier signal. This is achieved by redefining the amplitude modulation operation as follows: ye) All + mx(t)}e0s(Qot), (tay where m isa number chosen to ensure that (1+m.x(¢)] is positive for all 1. Figure 1.7 shows the waveforms ‘of a modulating sinusoidal signal of frequency 20 Hz and the amplitude-modulated carrier obtained ac cording to Eq. (1.14) for a carrier frequency of 400 Hz and m = 0.5. Note that the envelope of the ‘modulated carrier is essentially the waveform of the modulating signal. As bere the carrier is also presentChapter 1: Signals and Signal Processing Acos yt Hiker transformer Asin 2,1 Figure 18: Single-sideband modulation scheme employin as { : fae 1 au 7\ ne a =a, 0 a, Figure 1.9: Specira of pertinent signals in Figure 1.8. « modulate signal the proces is called simply deuble-sideband (DSB) od. carrier signal is Separated first and then used for demodslatacy ni cae ofthe conventional amplitude modulation, a canbe seen from Figure 1.4, the modulated capac wih of 225, whereas the bandwidth of the modulating esate eek ae mission medium, a motied form ofthe amplind ne in which citer the upper sideband oF the lower sidcben corresponding procedure is called single-side sideband modulation scheme of Figure 1 6a), or 2 to implement the single sdchand amplince modulation ig Hilbert transformer used is defined by Ea, (1.4), The spectra of pertne in Figure 19, ) modulation. At the receiving ’m- To increase the id of the modulated signal *hand (SSB) modulation to distinguish from the double- indicated in Figure 1.8, where the nt signals in Figure 1 8 are shown 1.25 Multiplexing and Demuttipiexing For an efciem utilization o sigmais are combined to fo Process of combining these: the XM 4 composite wideband sign signals is called mul arTON-bandwid led demutipexing ‘One widely used the frequenc method of combining division multiplexing band-limited to a low-frequency different voices (FM) scheme [ band of width 2225, ls in a telephone commu ication system is + each voice signal, typically 83), [Opp83}. Here, is freque ncy-transiated into a higher frequency1.2. Typical Signal Processing Operations " -2,° =a 0 a, a, a, (b) Figure 1.10; ustration of the frequency-division multiplexing operation. (a) Spectra of three low-frequency signals and (b) spectra of the modulated composite signal ‘band using the amplitude modulation method of Eq. (1.10). The carrier frequency of adjacent amplitude- modulated signals is separated by 25, with 2, > 22y to ensure that there is no overlap in the spectra of the individual modulated signals after they are added to form a baseband composite signal. This signal is then modulated onto the main carrier developing the FDM signal and transmitted, Figure |.10 illustrates the frequency-division multiplexing scheme. At the receiving end, the composite baseband signal is first derived from the FDM signal by demod ulation. Then each individual frequency-translated signal is first demultiplexed by passing the composite signal through a bandpass filter with a center frequency of identical value as that of the corresponding carrier frequency and a bandwidth slightly greater than 22,_. The output of the bandpass filter is then ‘demodulated using the method of Figure 1 6(b) to recover a scaled replica of the original voice signal. 1.2.6 Quadrature Amplitude Modulation ‘We observed earlier that DSB amplitude modulation is half as efficient as SSB amplitude modulation with regard to utilization of the spectrum. The quadrature amplitude modulation (QAM) method uses DSB ‘modulation to modulate two different signals so that they both occupy the same bandwidth; thus, QAM. takes up only as much bandwith as the SSB modulation method. To understand the basic idea behind the QAM approach, let; (t) and :r9(t) be two band-limited low-frequency signals with a bandwidth of 2, as indicated in Figure | 4(a). The two modulating signals are individually modulated by the two carrier signals A cox(S2y1) and A sin(Qot), respectively, and are summed, resulting in a composite signal y(¢) given by Y(t) = Axi (1) cOs(2p1) + Axa(t) sin( Qo) (us) Note that the two carrier signals have the same carrier frequency 2, but have a phase difference of 90° In general, the carrier Acos(¢t) is called the in-phase component, and the carrier A sin(S2et) is called the quadrature component. The spectrum Yj 12) of the composite signal y(1) is now given byChapter 1: Signals and Signal Processing cw = cos Nyt phase] @- oO OF phase se shiher } Lowpes L, a, Pas fiker ¥ 6) ) 1.11; Schematic representations ofthe quadrature amplitude modulation and demodulation schemes: dala YU) = $11 (2-24) + Xi 2 + 2, +4 (Xa U(2-2.)) - X24 and is seen to occupy the same bandwidth as: To recover the original modulatin, the quackature components ofthe +2,))} (1.16) the modulated signal obtained by a DSB modulation ignals, the composite signal is multiplied by both the in-phase and cartier 5 gnals: rately resulting in two U0) = yO) cos( ot). ralt) = y(t) sin(2r), (17) Substiuing »() ftom 89 (115) in Eg. (1.17), we obtain, afer some algcbra, U0) = $rrl0) + $y eos(2t + 4x0 sin’202, us) 12) = 40 + Srey sincrs ot) ~ Sxa(t)ons122,1), Lowpass filtering of r,(¢) and ra nals. Figure 1.11 shows the block demodulation schemes ) by filters with cutoff at 2m yields the two nodulating sig: Am representations of the quadrature amplitude 1 modulation and Asin the case of the DSB suppressed carrier mod ne at act epic of the caricr signal employ Ie in therefore not employed in the direct tansmnission of discrete-time data e-time segue ulation method, the QAM method also req ed in the transmitting end for accurate de ission of analog sig Seavences and in the transmission of analog =s by sampling and analog-to-digital conversion, res atthe discte Signals converted into 1.2.7 Signal Generation Ane ly important part o Signal generators is a devie integral partof the: It also has various ot There are # Signal processing is synthetic signal © generating & sinusoidal vPltade-modulaton and demodulation system de ther signal processing applications, plications that require the generation riangular waves. C simplest such Such a device is an signal, called scribed in the Previous two sections, of her types oF per : odie signals such as square iertain types o all fre fandom signals with a spectrum of constant amplitude for viomatan le white noise; often find applications rs Practice. One such application is in the fencration of cscrete-time synthetic speech sient,1.3. Examples of Typical Signals 13 | { ft A i i { ede @ 0 01 02 03 04 03 06 Seconds ) Figure 1.12: (a) A typical BCG trace and (b) one cycle of an ECG waveform. 1.3 Examples of Typical Signals’ To better understand the breadth of the signal processing task, we now examine a number of examples of some typical signals and their subsequent processing in typical applications Electrocardiography (ECG) Signal ‘The electrical activity of the heart is represented by the ECG signal [Sha8 1]. A typical ECG signal trace is shown in Figure 1.12(a). The ECG trace is essentially a periodic waveform. One cycle of the blood transfer process from the heart to the arteries is represented by one period of the ECG waveform as shown in Figure 1.12(b). ‘This part of the waveform is generated by an electrical impulse originating at the sinoatrial node in the right atrium of the heart. The impulse causes contraction of the atria, which forces the blood in each atrium to squeeze into its corresponding ventricle. The resulting signal is called the P.wave. The atrioventricular node delays the excitation impulse until the blood transfer from the atria to the ventricles is completed, resulting in the P-R interval of the ECG waveform. The excitation impulse then causes contraction of the ventricles, which squeezes the blood into the arteries. This generates the QRS part of the ECG waveform. During this phase, the atria are relaxed and filled with blood. The T-wave of the waveform represents the relaxation of the ventricles. The complete process is repeated periodically, generating the ECG trace. Each portion of the ECG waveform carries various types of information for the physician analyzing ‘patient's heart condition [Sha8!]. For example, the amplitude and timing of the P and QRS portions “This sexton hs been adapted from HandBiook for Dial Signal Processing, Sant K. Mitra and lames F Kaiser, Es. ©1993, John Wiley & Sons. Adapted by permission of John Wiley & SonsChapter 1: Signals and Signal Processing igure 1.13: Multiple EEG signal traces, indicate the condition ofthe cardiac muscle mass. Loss of amplitude indicates muscle damage, whereas plitude indicates abnormal heart rates. Too long a delay in the atrioventricular node is in dicated by 2 very long P-R interval. Likewise, blockage of some of all of the contraction i pulses is reflec! by imtermitent synchronization between the P- and QRS-waves. Most of these abaerretigs trees een with various drugs, and the effectiveness ofthe drugs can apain be monitored by obec i the new ECG waveforms taken after the drug treatment In practice, there are various types of externally produced artifacts theta {Fomé] Unless these interferences are removed, itis difficult fora physician to A 60-Ha power lines whose radiated el evpled io the ECG instrument throwgh capacitive coupli interference ae the electromyographic signals These and other interferences can be rer appear in the ECG signal make a correct diag lectric and magnetic fields are andlor magnetic induction. Other sources of Aare the potentials developed by contracting muscles. ‘moved with careful shielding and signal processing techniques. Electroencephalogram (EEG) Signal The summation of the elect the brain is represented by placed at va dif common source of noise is the ical activity caused by the random the EEG signal (Con), (Tom). In Positions on the scalp wi erences between the various electrodes a rom 0.5 to about 100 Hr, with the ampiit traces is shown in Figure 1.13, Both frequency-domain and time-domai epilepsy, sleep disorders, psychiatric ‘subelvided into the followin ‘ ‘multiple EE wo common electrodes placed on ue recorded. A typical bandwidth tudes ranging from 2 to. 100 m" recordings, the earlobes, and potential Of this type of EEG ranges NV. An example of multiple in analyses of the EEG signal have been use ‘malfunctions, and so forth, five bands: (1) the deita range, occupying ng the band from 4 19 & Hz; (3) the alpha ran ~oesupying the band from 13 10 22 Hz: and (5 for the diagnosis To this end, the EEG spectrum is ig the band from05 to 4 Hz; 2) the 8¢. occupying the band from 8 to 13 Hz; ) the gamina range, occupying the band1.3. Examples of Typical Signals 15 ‘The delta wave is normal in the EEG signals of children and sleeping adults. Since itis not common in alert adults, its presence indicates certain brain diseases. The theta wave is usually found in children even though it has been observed in alert adults. The alpha wave is common in all normal humans and is more pronounced in a relaxed and awake subject with closed eyes. Likewise, the beta activity is common in normal adults. The EEG exhibits rapid, low-voltage waves, called rapid eye- movement (REM) waves, in 4 subject dreaming during sleep. Otherwise, in a sleeping subject, the EEG contains burst of alpha-tike waves, called sleep spindles. The EEG of an epileptic patient exhibits various types of abnormalities, depending on the type of epilepsy that is caused by uncontrolled neural discharges. Seismic Signals Scismic signals are caused by the movement of rocks resulting from an earthquake. a volcanic eruption or an underground explosion [Bol93]. The ground movement generates elastic waves that propagate through the body of the earth in all directions from the source of movement. Three basic types of elastic waves ‘are generated by the earth movement. Two of these waves propagate through the body of the earth, one moving faster with respect to the other. The faster moving wave is called the primary ot P.wave, while the slower moving one is called the secondary or S-wave. The third type of wave is known as the surface wave, which moves along the ground surface. These seismic waves are converted into cleciical signals by a seismograph and are recorded on a sirip chart recorder or a magnetic tape Because of the three-dimensional nature of ground movement, a seismograph usually consists of three separate recording instruments that provide information about the movements in the two horizontal direc- tions and one vertical direction and develops three records as indicated in Figure 1.14. Each such record is a one-dimensional signal. From the recorded signals, it is possible to determine the magnitude of the ‘earthquake or nuclear explosion and the location of the source of the original earth movement. Seismic signals also play an important role in the geophysical exploration for oil and gas [Rob80}. In this type of application, linear arrays of seismic sources. such as high-energy explosives, are placed at regular intervals on the ground surface. The explosions cause seismic waves to propagate through the subsurface geological structures and reflect back to the surface from interfaces between geological strata, The reflected waves are converted into electrical signals by a composite array of geophones Isid out i patterns and displayed as a two-dimensional signal that is a function of time and space, gather, as indicated in Figure 1.15. Before these signals are analyzed, some preliminary time and amplitude corrections are made on the data to compensate for different physical phenomena From the corrected data, the time differences between reflected seismic signals are used to map structural deformations, whereas the amplitude changes usually indicate the presence of hydrocarbons, Speech Signals ‘The acoustic theory of speech production has led to a range of mathematical models for the representation Of speech signals. A speech signal is created by exciting the vocal tract using either quasi-periodic puffs of air or by creating turbulent air flow around a constriction in the vocal tract or by a mixture of these ‘two sound sources (Del93], [Rab78]. So-called voiced sounds are generated when air is forced through the tensed glottis, causing it to vibrate in an oscillatory manner and generating pseudo-periodic pulses of air that excite the vocal tract. Included in the class of voiced sounds are vowels such as /V/ (as in *big’) ‘rae (as in *bad"), voiced consonants such as /b/. (/ /g/./m/, in/ and so on; and so-called liquids and ‘slides such as (wi, /V, and /y/* are generated by forming a constriction at some point in the vocal tract, “The sounds ofa speech signal are wvally veprescnted picioilly wing 2 “phonetic alpaber” by inserting the marker “/* on oth sides ofthe eters representing the sound wih uppercase levers rpreseting vaced sounds that are stronger in amplitudesChapter 1: Signals and Signal Processing Nee Lt Chino Hits aneock reonded at sation Padiagsone Reser Southern California Seis Setect. Souhem Califoriaarthquake DataCenter. 29 uly Sve -chuip/iwwwalata scee one durations ofthe P-waves Approximate and S-waves have been aed tothe original seis ph which causes the air low to become turbulent (noise such asf), band so forth, Finally. there and hence has characteristics of both voiced 4 he woiced fricatives such as /v/,/x), and /aby There are sounds, called duction, such as fp) /V, and /k along the tract (thereby tot vocal trict). Pressure build Tike) and to act as the excitation source for sounds 8 & class of sounds that utilizes both sources of excitation nds and unvoiced sounds. Among this class of sounds are re sounds, whose For these ally blocking th ds up behind th removed. leading 1 a sound, ‘characteristics cha Sounds, the vocal tract hi 1 flow of ai ange dynamically during their pro- Ws a total constriction at some equivalently the production of sound i sudden release of pressure (nae BY a gradual release ofthe built-up air flow eure 1-16(a) depicts the speech waver point in the as the con. peech Dem 1 ey, mb titerance“ like digital signal processing * The ation of the speech waveform shown is 3s Magnified versions of Vin the word “like” aad the SY setment inthe word “processing” are sketches arying | in Figu ow-frequency voice, are evident from the mag, Periodic and can be modelo ind (c), respectively. The slowly -d waveform of 1 frequency unvoiced fricative 'eform in Figure 1.164) is of sinusoids. The lowest eMeY OF pitch, waveform of /S od by asum ofa finite number Presentation is called the fur lamentalfreque to be quasi frequency af oscillation Aequency. The unvoiced waveform in1.3. Examples of Typical Signals 17 Figure 1.15: A typical seismic signal trace gather. (Courtesy of Institute for Crustal Research, University of California, Santa Barbara,CA.) Figure 1.16(¢)has no regular fine structure and is more noise-like. ‘One of the major applications of digital signal processing techniques isin the general area of speech Processing. Problems in this area are usually divided into three groups: (1) speech analysis, (2) speech Synthesis, and (3) speech analysis and synthesis [Opp78]. Digital speech analysis methods are used in automatic speech recognition, speaker verification, and speaker identification, Applications of di speech synthesis techniques include reading machines for the automatic conversion of written text into speech and retrieval of data from computers in speech form by remote access through terminals or tle phones. One example belonging to the third group is voice scrambling for secure transmission. Speech data compression for an efficient use ofthe transmission medium is another example of the use of speech analysis followed by synthesis. A typical speech signal after conversion into a digital form contains about 64000 bits per second (bps). Depending on the desired quality of the synthesized speech, the original data can be compressed considerably; for example down to about 1000 bps. Musical Sound Signal The electronic synthesizer is an example of the use of modem signal processing techniques [Ler83]. [Moo77}. The natural sound generated by most musical instruments is generally produced by mechanica Vibrations caused by activating some form of oscillator that then causes other parts of the instrument to Vibrate. All these vibrations together in a single instrument generate the musical sound. In a violin, the primary oscillator is a stretched piece of string (cat gut). Its movement is caused by drawing a bow across its this sets the wooden body of the violin vibrating, which in turn sets up vibrations of the air inside 2 well as outside the instrument. In a piano, the primary oscillator is a stretched steel wire that is setPigure 1.16: Speech he let in “ike ES adatblenndbeel mh Chapter 1: Signals and Signal Processing [ie ty — Hn ath m example: (a) sen Pe: (a) setence-length segment, (b) magnif 4 segment (the letter () aged veson ofr so sin ofthe voiced segment in “processing”,1.3. Examples of Typical Signals 19 sme |e | cr a) ) Figure 1.17: Waveforms of (a) the cello and (b) the bass drum. (Reproduced with permission from J.A. Moores, Signal processing aspects of computer music: A survey, Proceedings of the IEEE, vol. 65, August 1977 pp. 1108-1137 ©1977 IEEE.) into vibratory motion by the hitting of a hammer, which in tar causes vibrations in the wooden body (ounding board) of the piano. In wind or brass instruments, the vibration occurs in a column of air, and a mechanical change in the length ofthe air column by means of valves or keys regulates the rate of Vibration, The sound of orchestral instruments can be clasified into two groups: quasi-periodic and aperiodic. Quasi-periodic sounds can be described by a sum of a finite number of sinusoids with independently varying amplitudes and frequencies. Figure 1.17(a) and (b) show the sound waveforms of two different instruments, the cello and the bass drum, respectively. In each figure, the top waveform is the plot of an éntire isolated note, whereas the bottom plot shows an expanded version ofa portion of the note: 10 ms for the cello and 80 ms forthe bass drum. The waveform of the note from a cello is seen to be quasi-periodic On the other hand, the bass drum waveform is clearly aperiodic, The tone of an orchestral instrun commonly divided into three segments called the ariack part, the steady-state part, and the decay part ‘igure 1.17 illustrates this division for the two tones. Note thatthe bass drum tone of Figure 1.17(b) shows no steady-state part. A reasonable approximation of many tones is obtained by splicing together these parts, However, high-fidelity reproduction requires a more complex mode! Time Series The signals described thus far are continuous functions with time as the independent variable. In many cases, the signals of interest are naturally discrete functions of the independent variables. Often such16 197 Month © (b) Figure 1.18: () Seasonally adjusted quan 1976 wo 1986 Gross National Product oft apt from (L491). (2) Monthly mean St.Louis, Mssou sary 1975 v0 1978 (adap fom (Mar) United States in 1982 doliars from temperature in degrees Ce ius For th jon. Examples of such sig of total monthly exports of a s are the yearly average number of sunspots, daily ‘country, the yearly population the annual yiclds per acre of crops in a country, and the monthly totals of Mional airline passengers over cerain periods. ‘This type of finite exient signal, usually called a rict,oxcars in business, economics, physical sciences, socialsciences, engi ical time series are shown in Figures 1.18(a fe many reasons for analyzing a particular time se to develop a mode! to determin. Variable and use it to forecast the future reasonably stock prices, the va i geograph eering, medicine, and ind 1.18(b) pli BoX70}. In some Ne the nature of the dependence of the data behavior of the series. As an e: accurate sales forecasts are necessary ‘components, and iti impe for predictin ations, there may on the independent ample, in business planning Some types of serie eran! fo extract these components. The study climate variations Fequire models based on their 8 Possess seasonal of periodic Of sunspot numbers is important invariably, the ime series data are noisy, and their representations statistical properties, earlier, an im ‘wo spatial varables,. Commot chest and dental X-rays dimensional signal for which the Variables and time. Figure 1.1 eS, radar and sonar images, three nction of three variables: two spatial fa digital i are image signal representation and modeling ions, analysis, and coding [Jai89) n physical quantity; a characterization of 4 photograph represents the luminances of n by a satellite oran airplane represents the the type of image and its applications, various dels are also based on perception and om local or ofthe image processing algorithms depend on the at any point is a fun a) shows the photograph o Problems in image processing io, reconstruction from project Each picture element i the element is called the Various objects 3s seen by th temperature profile of a The basic restr enhancement An infrared image taker Seographical area, Depending ‘ypes of image models are usually defined. Sect al characteristics. The natu image model being ased, re and performance g1.4. Typical Signal Processing Applications at Figure 1.19: (a) A digital image and (b) its contrast-eahanced version, (Reproduced with permission from Nonlinear Image Processing, S.K. Mitra and G, Sicuranza, Eds,, Academic Press, New York, ©2000 ) Image enhancement algorithms are used to emphasize specific image features to improve the quality of the image for visual perception or to aid in the analysis of the image for feature extraction. These include methods for contrast enhancement, edge detection, sharpening, lincar and nonlinear filtering, zooming, and noise removal. Figure 1,19(b) shows the contrasi-enhanced version of the image of Figure 1.19(a), developed using a nonlinear filter [Thu2000} The algorithms used for elimination or reduction of degradations in an image, such as blurring and ge- ‘ometric distortion caused by the imaging system andor its surroundings, are known as image restoration. Image reconstruction from projections involves the development of a two-dimensional image slice of a three-dimensional object from a number of planar projections obtained from various angles. By creating a ‘number of contiguous slices, a three-dimensional image giving an inside view of the object is developed. Image analysis methods are employed to develop a quantitative description and classification of one oor more desired objects in an image For digital processing, an image needs to be sampled and quantized using an analog-to-digital con- verter, A reasonable size digital image in its original form takes a considerable amount of memory space for storage. For example, an image of size 512 x 512 samples with 8-bit resolution per sample contains ‘over 2 million bits, dmage coding methods are used to reduce the total numberof bits in an image without any degradation in visual perception quality as in speech coding; for example, down to about 1 bit per sample on the average. 1.4 Typical Signal Processing Applications® ‘There are numerous applications of signal processing that we often encounter in our daily lives without being aware of them. Originally the signal processing algorithms used in these applications were carried ‘utinthe continuous-time domain. However, they are now being increasingly implemented using discrete signal processing algorithms. Due to space limitations. itis nt possible vo discuss all of these applications However, an overview of selected applications is presented “This section han ben adapted for Hondiook for Digital Signal Processing, Sanit K. Mien eal James F: Kaiser, Es, ©1993, John Wiley & Sons. Adapt by permission of Jon Wiley & SonsChapter 1: Signals and Signal Processing 401 compression <— Input in dB Variable threshold Figure 1.20 Transfer characteristic ofa typical compressor. 1.4.1. Sound Recording Applications The recording of most musical programs nowadays is usu ‘ound from cach instrument is picked up by its own micropho corded on a single n individu ‘made in an acoustically inert studio. The onc closely placed to the instrument and is track in a multitrack tape recorder containing as many a8 48 tracks," tracks in the master recording ar then edited and combined by the sound e system to develop a twortrack stereo reconting ick ty isthe clowenes ofeach individual microphone to its assigned insinment provides & senate orscarion terween he instrument and minimizes te backgreund nie in heecording cea, apart f one instrament canbe recorded ler if necessary Tit, dusing these don Avie, aha, nine can manipulate individ signals by using @ varity of sg ths musical balances between the sounds generated by the instrusnents eg ¢ signals sig There are a number of reasons for following I processing n change the 10 improve the quality of the transmi are (1) compressors and limiters, (2 lores H Limiter bende eaianles Tae ff times Fe] Bitns LY ewe | Bends Lo becsox J+ BH De > matt Detector -—> Figure 1.36: The tone detection scheme for TOUCH-TONE® dialing, filter _ #8 1.4.3 FM Stereo Applications For wireless transmission of a signal occupying a low-frequency ran 770 He 852 He 941 He 1209 He 1336 He 1663 He 31 Low group tones High group, tones such as an audio signal, it is ‘ecessary to transform the signal to a high-frequency range by modulating it onto a high-frequency carrier. At the re iver, the modulated signal is demodulated to recover the low-frequency signal. The signal Processing operations used for wireless transmission are modulation, demodulation, and filtering. TwoMusic Demo 1 32 Chapter 1: Signals and Signal Processing commonly used modulation schemes for radio are amplitude modulation (PM), We nent revi AM) and frequency modulation the basic idea behind the FM stereo broadcasting and reception scheme as used in the United States (Cou83]. An important feature of this scheme is that at the receiving end, the signal can be heard over a standard monaural FM radio with a st le speaker or over a siereo FM radio with two speakers. The system is based om the frequency vision multiplesing (FDM) ) method described earlier in Section 1.25, The block diagram representations of the FM stereo transmit Sf 37a) and (b), respectively At the transmiting end, the sum and the difference of the left and right channel audio signals, s(t) and s(t), are first formed, Note thatthe summed signal s,(1) + sed in monaural FM radio. The difference signal s(t) ~ s(t) is modulated usin suppressed carier(DSB-SC) scheme using a subcarrier frequency, modulated difference signal. and a 19-4H2 pi baseband signal sa(1). The spectrum of the signal is next modulated om ind the receiver are shown in Fig a(t) is he double-sideband foc Of 38 KHz. The summed tone signal are then added, developing the composite ‘somposite signal is shown in Figure 1.37(c). The baseband the main carrier frequency J using the frequency modulation method. At rece uing end, the FM signa is demodulated to derive the baseband signal sp (1), which we they sep- state nodulated difference signal, using cutoff frequency of the lowpass filter is around 15 ffeauency ofthe bandpass filter is at 38 kHz. The 19-KH pilot tone is wood 38-KHz reference signal for a cobcrent subcar The sum and difference of the 1wo © the low-frequency summed signal and the m ane a bandpass fier. The kHz, whereas the cemter in the receiver to develop the ‘demodulation for recovering the audio diff » signals create the desired left audic >and right audio signals. 1.4.4 Musical Sound Synthesis The generation of the sound of a musi another example of the a sical sound synthesis: and (4) physical mode a instrument using electronic circuits or software algorithms is rin of signal processing methods, There ar basically fout moose of oe ing title srthels. (2) spectral modeling symhess, 3) nonlnee oa hon Ming synthesis (Rab2001| {Sm Wavetable Synthesis In this widely used method, bes, are played bac playback Typical wh Prerecorded mu emory as databases called waveta- when needed. The stor tack and sustain portions. In the ] _analog. [9 lowpass input | ahold TT converter | | P filter | utput Figure 1.42: Scheme forthe digital processing of an analog signal structure whose parameters are adjusted using certain adaptation algorithms until the residual signal is satisfactorily minimized.” ‘Typically, an echo reduction of about 40 dB is considered satisfactory in practice. To liminate the problem generated when both subscribers are talking, the adaptation algorithm is disabled when the signal in the transmit path contains boih the echo and the signal generated by the speaker closer to the hybrid coil. 1.5 Why Digital Signal Processing?'' In some sense, the origin of digital signal processing techniques can be traced back to the seventeenth cen- tury when finite difference methods, numerical integration methods, and numerical interpolation methods ‘were developed to solve physical problems involving continuous variables and funetions. The more re- cent interest in digital signal processing arose in the 1950s with the availability of large digital computers. Initial applications were primarily concerned with the simulation of analog signal processing methods. Around the beginning of the 1960s, researchers began to consider digital signal processing as a separate field by itself. Since then, there have been significant and myriad developments and breakthroughs in both theory and applications of digital signal processing, Digital processing of an analog signal consists basically of three steps: conversion of the analog signal into a digital form; processing of the digital version; and finally conversion of the processed digital signal back into an analog form. Figure 1 42 shows the overall scheme in a block diagram form. ‘Since the amplitude of the analog input signal varies with time, a sample-and-hold (S/H) circuit is used first to sample the analog input at periodic intervals and hold the sampled value constant at the input of the analog-to-digital (A/D) converter to permit accurate digital conversion. The input to the A/D ‘converter isa staircase-type analog signal ifthe S/H circuit holds the sampled value until the next sampling instant. ‘The output of the A/D converter is a binary data stream that is next processed by the digital processor implementing the desired signal processing algorithm. ‘The output of the digital processor. ‘another binary data stream. is then converted into a saircase-type analog signal by the digital-ro-analog (DIA) converter. The lowpass filter at the output of the DYA converter then removes all undesired high- "Fora review of adaptive itering methods, se (C93) This scetion has been adept fro Handbook for Digital Signal Processing, Sanit K. Mita nd James F. Kaiser, Eds. ©1999, Jota Wiley & Sons. Adapted by permissica of John Wiley & SonsChapter 1 Time, in psec ) CHAE Time, in see , Ws Figure 143: Typical waveforms of signal ‘uipat ofthe SH circuit. e) AD ex analog output signal In (c) and (), te dg for clarty appears ie 142. (8) Analog input si PU ofthe digital processor. (e) D/A ‘converter output, and shown 38 positive and neyative pulses Amalog | Anaiog Processor Analog inper output Figure 144: Analog processiny of analog signals frequency components and detive the waveforms ofthe pert levels ofthe binary signals In contrast to the ena itt te desired processed analog signal, igure 1.43 illustrates als at various stages in the Process, where for © shown as Zative pulse, respectively shove, a direct analog nce it imvlves ony a single the advantages are of There are of cou ty the two igital processing of Se many advantages iscused next (Be!2000), Prog} Unlike ‘operation of di signals. AS a result, a digital independentof iemperat easily in he most important ones ‘analog circuits, the, ital circuits do Sircuit is less sensitive to x nd does er, it is amenable to Des not depend on precise Values of the digital ole inces of component values and is fairly ital circuit can be reproduced onstruction or later whi ances in very large scale lume quantities an in use. Mor Scither dui ‘on, and with the recent ad1.5. Why Digital Signal Processing? 39 of «4 6 + ee ow ow Time, in jt see Time inj see © Figure 1.45: Illustration of the time-sharing concept. The signal shown in (c) has been obt time multiplexing the signals shown in (a) and (b). integrated (VLSI) circuits, it has been possible to integrate highly sophisticated and complex digital signal processing systems on.a single chip. In a digital processor, the signals and the coefficients describing the processing operation are repre sented as binary words, Thus any desirable accuracy can be achieved by simply increasing the wordlength, subject to cos’ limitation. Moreover, the dynamic ranges for signals and coefficients can be increased still further by using floating-point arithmetic if necessary Digital processing allows the sharing ofa given processor among a number of signals by timesharing, thus reducing the cost of processing per signal, Figure 1 45 illustrates the concept of timesharing, where ‘wo digital signals are combined into one by time-division multiplexing. The multiplexed signal can then be fed into a single processor. By switching the processor coefficients prior to the arrival of each signal at the input of the processor, the processor can be made to look like two different systems. Finally, by demultiplexing the output of the processor, the processed signals can be separated. Digital implementation permits easy adjustment of processor characteristics during processing, such 4s that needed in implementing adaptive filters. Such adjustments can be simply carried out by periodically changing the coefficients of the algorithm representing the processor characteristics. Another application ‘of the changing of coefficients is in the realization of systems with programmable characteristics, such as frequency selective filters with adjustable cutoff frequencies. Filter banks with guaranteed complementary frequency response characteristics are easily implemented in digital form Digital implementation allows the realization of certain characteristics not possible with analog imple- ‘mentation, such as exact linear phase and multirate processing. Digital circuits can be cascaded without any loading problems, unlike analog circuits. Digital signals can be stored almost indefinitely without anyChapter 1: Signals and Signal Processing 40 loss of infomation on various storage medi, such ss magnetic tapes and disks and optical disks. Such sored sigs cn ler be roceed oe uch ain he cmpec clk pl player the digital audiotape payer, o simply by usin he other hd, store al video disk a general purpose computer as in seismic data analog signals deteriorate rapidly as time progresses and cannot be processing. On tage isthe applicability n seismic applications, whe physically very large digital processing to ver induct low frequency signals, such s and capacitors needed for ana so associated with some disadvanta; the increased system complexity in the digital processing of analog signals because ofthe nced for addi tional pre-anxd postprocessing devices such as the A/D and D/A converters and their associated filters and complex digital cireitry A second disadvantag s. One obvious disadvantage is associated with digital si l processing is the limited range of frequencies property limits its application particularly in the digital processing of op PP Ps eral, an analog continuous-time signal must be least twice the highest Frequency com, available for processing. Thi log signals. As shown later, i that is sampled at a frequency }ponent present inthe signal- If this condition is not satisfied, va components with fequencics above hal the simpling frequency appeas as signal components below this particular fi put analog signal waveform. The available fi ‘digital signal processor is primarily determined by the S/H circuit and the ND convererand.as esl, isLimited by the state of the at of the technology, the highest sampling ueney ported inthe literature presenily is around | GHz [PouS7]. Such highs ‘ot usually used in practice since the achievable the distal equivalent ofthe analog sample example, the reported resolution of hand, in most application, - quency, touily distorting the (quency range of operation of sampling frequencies are resolution of the A/D converter, given by the wordlength decreases with an increase in the speed of the converter. For # an A/D converter operating at 1 GHz is 6 bits (Pou87}, On the other the required resolution ofan A/D converter is from 12 bits to around 16 bite Tae ano most 10 MEz is presenily a practical upper limit. This upper ‘and larger with advances in technology The third disadvantage stems from the fact thar that consume electrical po over 405,000 algorithms digital systems are constructe or rt Example. the WE DSP32C Digital Signal Process chip contains Aisistors and dissipates around 1 watt. On the other hand, 4 Variety of analog processing sin; implemented using passive circuits employing inductor. capacitors, and resistors that 4o not need power. Moreover, active devices ae les ble than passive components. Aiother dadvantage of digital signal processing is dc ie the effects resulting fi tvcneawih oic pecsion aitmeticn nee ee aorta. By poy and its implementation, the effects of the 2 d using active devices from the algorithms designing the algorithm nite wordlength can be minimized in most cases iiomever the advantages far outweigh the disadvareess in various applications, and with the con tinuing dec tidy 2% OF ditial processor hantware. applications of digi increasing rapidly al signal processing areChapter 2 Discrete-Time Signals in the Time Domain As indicated in Figure 1.1(¢), a discrete-time signal in its most basic form is defined at equally spaced discrete values of time, the independent variable, with the signal amplitude at these discrete times being 4 continuous variable. Consequently. a discrete-time signal can be represented 2s a sequence of numbers, ‘with the independent time variable represented as an integer in the range from —o0 to +00. Discrete-time signal processing, then, involves the processing of such a signal by a discrete-time system to develop another discrete-time signal with more desirable properties or to extract certain information about the original signal In many applications involving continuous-time signals, itis becoming increasingly more attrac: tive to process the continuous-time signal by discrete-time signal processing methods. To this end, the ‘continuous-time signal is first converted into an “equivalent” discrete-time signal by periodic sampling; the resulting signal then is processed by a discrete-time system to generate another discrete-time signal, ‘andthe latter is converted into an equivalent continuous-time signal, if necessary. As we shal show lat Section 3.8.1 ,under certain (ideal) conditions, the conversion of a continuous-time signal can be carried ‘out such that the discrete-time equivalent has all the information contained in the original continuous- time signal and, if necessary, can be converted back into the original continuous-time signal without any distortion, ‘Thus, to understand the theory of digital signal processing and the design of discrete-time systems, we need to know how to mathematically represent discrete-time signals and systems. Such representations can be either in the time domain or in the frequency domain. In this chapter and the following chapter we restrict our attention to discrete-time signals, In this chapter, we consider the mathematical representations of discrete-time signals in the time do: main and several concepts associated with such representations. We first consider the representation in the time domain. We then describe several elementary operations on discrete-time signals and the de- Vices used to implement these operations. These operations form the basic building blocks in developing discrete-time systems for the processing of discrete-time signals. We next discuss various classifica of the discrete-time signals that often can be used to simplify the signal processing algorithms. A. num: ber of basic operations that generate ether sequences from one or more sequences are described next. We then describe several basic discrete-time signals or sequences that play important roles in the time domain characterization of arbitrary discrete-time signals and discrete-time systems. Next we discuss the relation between a continuous-time signal and its discrete-time version, obtained by sampling the former at uniform intervals. In addition, we point out the condition under which the discrete-time signal obtained by uniform sampling can uniquely represent the parent continuous-time signal. One type of similarity 4Chapter 2: Discrete-Time Signals in the Time Domain te utl| ; Figure 2.1: Graphical representation ofa discrete-time sequence measure between a pai of discret roduced e-time signals is given by the cross-correl lation sequence which then is ind some ofits properties ‘re investigated. Finally the chapter concludes with a qualitative ne random signals, this chapter and successive through computer simulations, the v chaplers, we make extensive use of MATLAB to illustrat, rarious concepts introduced. 2.1 Time-Domain Representation aearie ue cater, in digital signal processing signals ae presented as sequences of numbers called ineee gecaile value of atypical ciscete-tre signal or sequence is a finction of te independent tin local ffm defined inthe range ~co and oo. For example, xf isthe amplitude of a discrete fore ila 3 the time instant x where 00 < n < 00. should be noned sha x (| is defined only degen es of and is undefined fr noninteger values ofn, Ifa dscns tin Presets Of mumbers inside braces, the location of the sample ' indicated by an arrow + under it. The sample values 1¢ te SERIE values its eft are for negative values fn. An examph samples is given by signal is written as a alue associated with the time index ight are for positive values of m, and the Ic ofa discrete-time signal with real-valued 9.95,-02,2.17,1.1,02,-3.67,2.9,-08.4.1,..} 1) Forthe above signa, x{-1] = ~02,x(o) = 2 17x Sseavence {ln} with real-valued samples is illustrated i |i some applications a discret 1.1,and so on, The. ‘2raphical representation of in Figure 2.1 serene (xi is generated by periodically sampling a continuous- time signal x4(¢) at uniform time interval ‘ acc X11 = allan = ...,-2,-1,0,1,2, 2) 4 illustrated in Figure 22 sampling interval ot samp! Samples in Eq. (2.2) is called the the sampling pains interval 7’. denoted as Fr, is called he 23) T 7 ax ls pet second, oe Here (Hf the samp ‘case of an image, a sp ates2.1. Time-Domain Representation 43 Figure 2.3: A digital signal It should be noted that, whether or not a sequence {jn} has been obtained by sampling, the quantity xn] is called the nth sample of the sequence. For a sequence {x(n}, the nth sample value xn] can, general, take any real or complex value. If x(n} is real for all values of 1, then {x{n)} isa real sequence On the other hand, if the nth sample value is complex for one or more valves of m, then it is a complex sequence, By separating the real and imaginary parts of xn], we can write a complex seque fxfa}} = (rela) + 7 (xinlnT} @a) ‘whete en] and rigfi] are the real and the imaginary parts of x/n]. respectively, and are real sequences. For a real sequence, Xj] = 0, and for a purely imaginary sequence, x,[n] = 0, for all values of 1, The complex conjugate sequence of {x{1}} is usually denoted by {x*[n]} and waitten as {x*fnl} = tell} ~ j (xmlt]}-! Often the braces are ignored to denote a sequence if there is no ambiguity ‘As defined in the previous chapter, there are hasically (wo types of discrete-time signals: sampled data signals, in which the samples are continuous-valued; and digital signals, in which the samples are discrote-valued. The pertinent signals in a practical digital signal processing system are digital signals obtained by quantizing the sample values either by rounding or by truncation. For example, the digital signal {(n]} obtained by rounding the sample values of the discrete-time sequence «| of Eg. (2.1) to the nearest integer values is given by (lol) = foo 1,0, 2, 1,0, 43, 14 t Figure 23 shows a digizal signal with amplitudes taking discrewe integer values inthe range from —3 to 3 For digital processing of a continuous-time signal, it is first converted into an equivalent digital sig. ‘al by means of a sample-and-hold circuit, followed by an analog-to-digital converter. The processed "The complex contion operation i dented hy te superscrp symtol *Chapter 2: Discrete-Time Signals in the Time Domain all... (b) alef-sided sequenc igure 24: (a) A right-sided sequence ‘igital signal is then converted back into an equivalent continuous-time signal by a digital-t Verte followed by an analog reconstruction filter. Section 3.8 is concerned with the d Siuminuous-time signals It develops the mathematical foundation of the sampling process and describes the ns of various interface circuits bet ‘Chapter 12 considers the effect of discretz ital processing of tween the continuous-time domain and ion of the armplitudes. he digi I domain, 2.4.1 Length of a Discrete-Time Signal The discrete-time signal, wy be a finite-lengeh or an infinite-length sequence fiite-duration or fnte-extent) sequence is defined A finite-length (also called d only fora fini NM; =<, weer ~20 < Mi and Ni < oo with Ny > Nj. The length or deration N of the above finite-tength N=N M+ 2.6) ‘ A length-N discret time sequence consis ence also can be samples whose arguments are outs sof N samples ands often referred to ‘considered as an infinit-length sequ le the above range. The process of le ppending with sero Sequence by append orm < Ni: that is, 2 an N-point sequence A tinite-length seq Point seq nce by assigning zero values to "ngthening a sequence by adding or cerv-padding. In some applications, a finite-length ing it with non-zero samples ero-valued samples is called ap nce is also made a longer There are *- A righ-sided sequence x(n) has zero-valued sam x] =0 — forn < Ny, te integer that can be led a causal sequence? Likewise Positive or negative. IF Ny > 8 right-sided sequence usually is 8 lefi-sided sequence x \n} kas 70 valued samples for > Na: that *]=0 — forn > Ny be positive where Ns is finite integer that called an antcausal sequence OF negative. If Np S 0, a left-sided sequence ust moa eal mo-sided sequences defined for bah painnen nega Fee simpy ig taste abv two types of ene sed sequences i ‘rans n fake length sequences defined for positive valure the time index m bevinning at iniccn moan i ene Leet Wil be assumed tobe the one esrcint ee et wine ing explicitly shown under it domain. As we shal! show in Chapter 5, 9 finite-length eo ee an Pera eam also be represented as a a te-time signal Oa i rig I i camality cond2.1. Time-Domain Representation 45 finite-length sequence in a transform domain where the sample of the sequence is a function of a discrete variable in the transform domain, usually denoted by the integer variable k. It is a common practice to denote the transform domain representation of a discrete-time signal x[n] with capital letters; for examp X[k], If we denote the vector of the time-domain samples of length N’ defined for loo [sn 2.10) It follows from Eq.(2.9)that fora length-.V sequence, | x |12 / VN isthe root-mean-squared (rms) value of {xf}, and |x Ji /V, is the mean absolute value of {x[a}}. It can be shown that (Problem 2.2) Ixibsixjh. One application of the norm isin estimating the error inthe approximation of a discrete-time signal by ‘another discrete-time signal. Lethe length-N sequence yi], < 1 < of the length-N sequence x(n],0 *., dy) become very smalll due to the randomness of the noise. Example 2.1 illustrates ensemble averaging. EXAMPLE 2.1 Hlustration of Ensemble Averaging ‘We assume for simplicity that the original uncorrupted data is given by [Sch75] sn) = 21010.9)"] 216) ‘The MATLAB program to gencrate the above data s{n}, the noise d[n], and the ensemble average are given in Program 2.1. The two sequences generated using the above program are shown in Figure 2.6, ‘One sample of the noise-corrupted data s}1] + dj fn] and the ensemble sverage obtained after SO mea- surements is shown in Figure 2.7. It can be seen thatthe ensemble average shown in Figure 2.7(b) is nearly the same 28 the original uncorupted data shown in Figure 2.6(a). Program 2 1.m ‘An application of ensemble averaging is in the power spectrum estimation of a random signal, dis- cussed in the CD accompanying this book. ‘The time-shifting operation illustrated below in Eq. (2.17) shows the relation between xj] and its time-shifted version w4[1} xn—N}. 17)Chapter 2: Discrete-Time Signals in the Time Domain AUD eA (b) Figure 24: (a) The orginal uncorrupted seat ce sn} and (b) the noise sequence dyn ed sequence Ensemble average In nt o) Figure 2 ‘ sample ofthe corrupted sequence and (b) the ensemble a © after 50 measurements, Fond asiiteper IC > 0, itis a delaying operation, am ForN = 1, we have if N < 0, itis anad the input-output relation dvancing operation. wala] = xn 1, fand the device implementing the del lay operation by one transform. introduced later in Cha sample period is called a unit pter 6, the above ay. In terms of ‘elation can be rewritien as, Wale) = 21K (2), Foe eee, tmntforns of the otpat sequence inn} and the input Iisa usual practice 1 repre } unit delay operation using the symbol shown in Figure 25 toe The opposite ofthe un 4(2) and X(2) are, res m SEI operation is the uit vance operation, defined by “hich interme ofthe z-transform is given by Ws( 2 X(2)2.2. Operations on Sequences 49 AAs a result, the unit advance operation commonly is represented schematically using the symbol 2, as shown in Figure 2.5(e) The rime-reversal operation, also called the folding operation, is another useful scheme to develop a ew sequence. An example is weln} = xl-n), (2.18) which is the time-reversed version of the sequence x] In Figure 2.5(f), we show a pick-off node, which is used to feed a sequence to different parts of a discrete-time system. Example 22 illustrate the use of three basic operations: product, addition, and multiplication EXAMPLE22 Basie Operations on Sequences of Equal Lengths Consider the following two sequences of length 5 defined for 0 Sveral new sequences of length $ generated fom the above sequences are given by = elu) dal = (544, ela) +4 bn] = (4.9, 40 wale] = Jeln] = (112. 1435, 126, As indicated by Example 2.2, operations on two or more sequences to generate a new sequence can be carried out if all sequences are of the same length and defined for the same range of the time index n However, ifthe sequences are not of equal length, all sequences can be made to have the same range of the time index n by appending zero-valued samples tothe sequence(s) of smaller lengths. This process is illustrated in Example EXAMPLE 2.3 Basic Operations on Sequences of Unequal Lengths Consider a sequence {en}} of length 3 defined for 0 stn kau) (2.200 Uz time change of vaables, povided th conta, an be seen from th aoe m sum given in Eq. ( n of the sequence y 202) is convergent. AS "© equations, the generation te coche {fn} by the convo2.2. Operations on Sequences 51 involves four basic operations: time-reversal, multiplication, addition, and delay. We shall show later in Section 4.4.1 that a certain class of discrete-time system is characterized completely in the time domain by the convolution sur For brevity, the convolution am expression will be compactly written as vin] = x{n] @ Ala), @2n where the notation @ denotes the convolution sum. Example 25 il ‘convolution sum. strates the computation of the EXAMPLE2.5 —Mlustration of Convolution Sum Computation ‘We systematically develop the sequence [1] generated by the convolution of the two finite-length sequences x{n] and Al} shown in Figure 2.9(a). As can be seen from the plot of the shifted time reversed version {A[n—£]} for n < 0 sketched in Figure 2.9) for n = —3, the sample index &. the kth sample of either {x{k]} or {h(n —K]} is zero, As a result, the product ofthe Ath samples is always zem for any k ,and consequently, the convolution sum of Eq, (2.20) leads forn <0, ‘Consider now the calculation of y[0}. We form {h{—k]} as shown in Figure 2.9(c). The product sequence {s[&é{-A}} is ploted in Figure 2.94), which has a single nonzero sample for k = 0, x{O)A(0). Thus, {0} = s(O}Alo) = 2. For the computation of 1), we shift {h[~A} to the right by one sample period to form {h[1 — k]},as stetched in Figure 2.(¢). The product sequence {{kJh{! |} shown in Figure 20(7) has one nonzero sample for k = 0, x{O)A{l). As result, 4405-4, It] = xlOIAlt] + xA)AQ0) To calculate y2}, we form {h[2 — k]}. 2s indicated in Figure 2,9%g). The resulting product sequet {rIKJAL2 — &]} i plotted in Figure 2.9(h), which leads to YEO) = (OVAL) + xLA(1] + xL2AI0} = 04041 = 1 ‘This process is coatinved, resulting in xls + =(1)A02 + xPIALN] + a13(0) = 24042-1253. yi] = x(1WE3) + x22] + xBIAUH] +als}4(0) = 0+0-2+3~=1 (2D) + xD] +-x18)A[1] = -1404 6= 5, HIS] + x19) v1] = sla} = 3. From the plot of /ifa —A]} form > 7 as shown in Figure 2.9() and the plot of {x(&}) in Figure ican be seem that there is no overlap between these two sequences. Hence. yk The sequence {y[n]} as obtained above is sketched in Figure 2.10.2.2. Operations on Sequences 53 vin Amplitude Time index n Figure 2.11; Sequence generated by convelution wsing MATLAB, It should be noted that the sum of the indices of each sample product inside the summation of either Ey, (2.20a) or Eq. (2.206) is equal to the index of the sample being generated by the convolution sum operation. For example, the sum in the computation of y(3} in the above example involves the products <(0)h3}, x{1)(2}, x{2)(1}, and x{3}A(0]. The sum of the indices in each of these four products 03, which is the index of the sample y[3]. As can be seen from Example 2.5, the convolution of two finite-length sequences results in a finite- length sequence. In this example, the convolution of a sequence {x(n} of length 5 with a sequence {hn} of length 4 resulted in a sequence {y[n]} of length 8. In general, ifthe lengths of the two sequences being convolved are M and NY, then the resulting sequence after convolution is of length Mf +N — 1 (Problem 2.5) In MATLAB, the M-file conv implements the convolution of two finite-length sequences, tion is illustrated in Example 2.6 equal applica EXAMPLE26 — Convolution Computation Using MATLAB Program 2.2 given in the accompanying CD can be used to compute the convolution of two finite- length sequences. The plot of the sequence generated by convolving the two sequences of Example 2.5 lasing Program 2 2 is indicated in Figure 2.11. As expected, the result is identical to that derived in Example 25, “in the inerature, the symbol forth convolution sum i + without the circle. However, asthe supecscrigt * i always wscd for coring the complex coniuatin cperatin. inthis text we have adopted the symbol @ to dence the convaation sum operation. Program 2,54 Chapter 2: Discrete-Time Signals in the Time Domain wi Ife Loan a Afr Lose (b) @ Figure 2.12: Represen fon of basi sampling rate alteration devices: (a) up-sampler and (b) down-sampe 2.24 Sampling Rate Alteration Another quite useful operation is the sampling rate alteration that ‘with a sampling rate higher or lower than that sampling isemployed t rate a new sequence a given sequence. Thus, if x{n] is a sequence with a 1 Fr He and itis used to generate another sequence [1] with a desired sampl Fj He, thea the sampling rate alteration ratio is given by rate of fr 22 Rit [hs cperation for R > 1 results in a sequence with a higher sampling rate, This operation is used in ree ene & sequence with slower sampling rate 10 a Sequence with a higher sampling rate. The reg te tine system implementing the interpolation process is called an interpolonor On the other hand, if R <1, the samy ed by & process called decimation. The discrete-time system implementing the decimation process is called a decimator The basic ope ations employed in the sampling rate alte on process are called up-sampling and crate crea. Tes cpcrations pay important roles in maltrate discrete ime system adne once cred in Chapters 13 and 14, ENE > 1.1 1 epic sort cages ae nen by the up-sampler between each two con ecutive samples of the input sequence x(n] to develop an output sequence x, n] according to the relatin x(n} = {*h/L), n=0.41,492,..., RPG otherwise , is L times large than that of the presentation ofthe up-samph The sampling rate of xy a} The block diagram re Original sequence x(n} in Figure 21 degra pen also called a sampling rate expander, is shown “Ad, Figure 2.13 illustrates the up-sampling operation for ag ip-sampling factor of L = 3. up-sampling factor o Te, Peltor consists ofan upsumpler flowed by Sitere on system that replaces the inserted Ta Aa ames of xn] with more appropiate vas omer "le. A simple example of sn iterpolators given in Sorta Fi eth down-sampling operation by an integer fo sping every Mt sample of xn and removing An betweer ln] according to the relation ‘ 4 linear combination of samples of > 1 on a sequence xn} consists of 'n samples, generating an output sequence vla] = xjnaty ona in Asquene Jp] whose sampling ‘with indices equal to an inte f Mt The schematic represent ure 2.12(b). Figure 2 (224) te is (1 Bet multiple of M are tation of the 14 illustrates the dex DF Consists of a discre M)-th that of x(7). Ba Fetained atthe output, and all down:sampler of samp rn-sampling operation asically, all input samples others are discarded Hing rate compressor is shown in Fig. for a down-sampling factor of M = 3. ne system followed by a do a pwn-sampler, The discrete-time system iscaused by the down ee ai She! hn is appropriately band-limited 10 prev . Ing Operation. down-sampler ensures t2.3, Operations on Finite-Length Sequences 55 Input sequence Ouput sequence upsampled by 3 1 f ' stip )_GR) 18 | 0 10 2% 30 40 50 Time index » @ (b) Figure 2.13: Hlustration ofthe up-sampling process. Input sequence ‘Output sequence downsampled by 3 o 0 2 » 40 30 o 0 2% 3% 4 50 Time index n ‘Time index m @) () Figure 2.14: Illustration of the down-sampling process. 2.3. Operations on Finite-Length Sequences Some of the operations on sequences described in the previous section are not applicable to finite-length sequences, Let xn] be such a sequence of length N defined for 0 N.. A time-teversal operation on x} results in a length-N sequence x{-n] defined for —(N ~ 1) < n <0. Likewise, linear time-shift of the sequence by an integer valued ‘M results in a length-N sequence x{n + M], which no longer is defined for 0 0, the above equation 0, it isa right circular shift, and if ny <0, itis a left plies xn —no}, forny DL bled? (240) EXAMPLE 210 Example of a Power Signal Consider the causal sequence defined by It follows from Ei, (2.34) that fr] has infinite energy. On the ether hand, from Eq. (2.37), is a power is piven by Which is finite,62 Chapter 2: Discrete-Time Signals in the Time Domain Other Types of Classification A sequence xr] is suid to be bounded if each of is samples is of magnitude less t positive number By that is, i oF equal to a finite ipaebis ap igure 2.19 is a bounded sequence with a bound B, = 2. A sequence x] is said tobe absolutly sunumable if D bbe <2 o The periodic sequence of Fi A sequence is said to be square-summableif ¥ beh? <0o 243) squence, therefore, has finite ene Of a sequence tha is square. A square-summable se An example "ay and is an energy signal if it also has zero power. summable but nt absolutely summable is at, condo, shl= [are <0 Examples of sequences that are neither absolutely surmmable or square-summable are whl =sinen, — —o9 xta-+ kn 44) where N is a posi = 9 inte is a periodic sequence *espence hn i called an N peri extension f+ With @ petiod V (Problem 2.25). The periodic ‘We now consider se time systems. For ¢ Snportanrolesin the analysis and design of discrete ‘d in terms of some of these time signal processing is the basic Representation of a cla sequences arbitrary i 88 of discrete-time syste ystems in terms of the resper This representation permits th pa iscrete-time signals ifthe 24.1 Some Basic Sequences The most common bas sequence, an PORN ee the un sempe sequence, thy wwence, and the exponential sequence These sequet rsa nces are defined next. mae oats ida2.4. Typical Sequences and Sequence Representation 63 T33T3s SSSI 23 tss @) ©) Figure 220: (2) The unit sample sequence {8{n]} and (b) the shifted unit sample sequence {6h — 2}. @ (b) Figure 2.21; (a) The unit step sequence {[n]} and the shifted unit step sequence {1 (4 +2)} Unit Sample Sequence ‘The simplest and one of the most useful sequences is the unit sample sequence, often called the discrete time impulse or the wnit impulse, as shown in Figure 2,20(a). It is denoted by 8[n] and defined by 1, n=0, ainl={o. 020 (24s) ‘The unit sample sequence shifted by k samples is thus given by tine, nek Figure 2.20(b) shows dfn — 2]. We shall show in Section 2.43 that any arbitrary sequence can be rep resented as a sum of weighted time-shifted unit sample sequences. In Section 3.4, we demonstrate that ‘@ certain class of discrete-time systems is completely characterized in the time domain by its output re sponse to a unit impulse input. Furthermore, knowing this particular response of the system, we can compute its respoase to any arbitrary input sequence. Unit Step Sequence A second basic sequence is the unit step sequence, shown in Figure 2.21(a). It is denoted by sefn] and is defined by 1 n=O. wel= {0 t20 246) The unit step sequence shifted by k samples is thus given by fl nek lo. n fla — mj = > ate 2.478} Sin] = ulm] — ula —1), (2.47) Sinusoidal and Exponential Sequences A commonly encountered sequence is the re sinusoidal sequence with constant amplitude of the form 4 [0] = Acoslenn +9), 20 0), or decayin is the amplitude of the si: ental sequence are real ing (¢o < 0) amplitudes for2.4. Typical Sequences and Sequence Representation Time index n 1 0 con nore rithll < | decays exponentialy an increas sa le] > 1 grows exponentially 224. ames of real exponential sequenees obtained fortwo sel tg 224 We shall show later in Section 3 ‘complex exponential sequer The re with , = 0 N= complex ex amplitude. Note that in the With both 4 and are shown in Figure 2.1 that a large ass of sequer nes of the form e: Wes can be expressed in terms of ‘iausoidal sequence of Eq. (2.48) and the Periodic sequ 2er, where Nan s the fundamental period complex exponential sequence of Eq. (2.528) aces of period Na long aso, N is an integer multiple of 27r; that is, “are positive integers. The small lest possible N satisfying this condition sey omtene®. To verity this, consider tae to nen sinusoidal sequences “ale] = costonn + $) and sofa] = costa, (n + N) + @). Now In} = cos(onin +N) 4.4) ~ Ny pplication of the aboy h or very long leagth sequence is in extrac all samples in the range 1 << of with tal a and Sting other samples outside the aac values by multiplying [a] with wa]. Thon Spemeteesst [% n «2 >> 0 > 0. The product sequ yb1] = xi balx2[a] is then given by os{aoxn) costo) 258 Using a trigonomettc identity, we ative at vn] = $.com((on + o2)n) + on) es») “The new sequence y[n] is thus composed of to sinuscidal sequences of frequencies 0, + «2 and @2 ~ a. It Should be noted that, because of the property given by Eq. (2.55), if wy + aoa >. the‘Chapter 2: Discrete-Time Signals in the Time Domain of to sas sequen Waveform of prods oft sno ¥ , Time index Flere 2.26; Prats ftw sini sequcacs with anptr foquenics oy = 01x andon = 0.01" bigh-freqcncy sinusoidal sequence cos((oy +o) ay nM) with an angular frequency of va ars 28a low fre le smaller than x rency sinusoidal sequence Figure2.26 showsthe plot ofthe seque 1 = O.17 and o = 0.012, respective product signal is a hig ata lower Frequency that isa pro Asc frequency sinuscidl so oduct of two sinusoidal sequences of frequencies an be seen from this figure and also from Ea. (2:58), the ‘avenee with an amplitude sinusoidally varying with time Fundamental and Harmonic Components It should be noted thatthe a ln) and the an whose angul called the famponent with its frequency being called the frequency that kines that ofthe Fundemeng the sequence x; isthe funda “Y of the sequence x2(n] in Example 2.13 is 3 mes that of x[n). The sinusoidal sequences quence of lower an *juence with the lower frequency i cal fundamental frequency. The harmon alifequency i called the k-th harmonic. In Example 2.13, mental component with a fundamental angular frequency of 0.05 radian and fn), ae respectively, the third and an harmonics the form of line es that of gular frequency are led the fundamental ie component with a As shown in Example 2.14 angular frequencies « and soda signals x, [n) a angular frequencies are sum, id x2[n] with nor ect of two sinusoidal signals whose normalized “ sof he ena at 22s espectively thats the rane nd differences of the normal ted angular frequencics ofthe orig Sinusoidal signals. Uw» <& ost the two normalized angular wing (0 Secon 25) C0 + O10) st tecacy sien sem((Qm ~ ny — era) i called2.4. Typical Sequences and Sequence Representation n Ls Figure 2.27: An arbitrary sequence x(n] frequencies w +2 and ©; — e are nearly equal to each other, and, as pointed out in Figure ‘multiplication process generates a high-frequency signal, called a beat nore, which fades in and out at a rate determined by the lower frequency. Ifthe two frequencies are in the audible range, the beating effect, ccan be heard. An application of the beat signal generation is in the tuning of two musical instruments, ‘The beat note cannot be heard when both instruments have the same pitch. 2.4.2 Sequence Generation Using MatLas MATLAB includes a number of functions that can be used for signal generation. Some of these functions of interest are exp, sin, cos, square, sawtooth For example, the code fragments to generate a length-N’ complex exponential sequence with an exponent a + jb and of the form shown in Figure 2.23 is given by n= 1:N3 X = Keexp((a + bei)en); The complete code is given in Program 2.3. Likewise, the code fragments to generate a length-(N + 1) real exponential sequence with an exponent a and of the form shown in Figure 2.24 is given by The complete code is given in Program 2.4, Another type of Sequence generation using MATLAB can be found earlier in Example 2.1 2.4.3 Representation of an Arbitrary Sequence An arbitrary sequence can be represented in the time domain as a weighted sum of the delayed and advanced versions of a basic sequence. A commonly used basic sequence in the representation of an arbitrary sequence is the unit sample sequence. For example, the sequence x{n] of Figure 2.27 can be expressed as x(n] = 0.58fn +2) + 1.55)n — 1] — [a 2] + 8fn — 4) + 0.75 8[n — 6), (2.60) An implication of this type of representation is considered in Section 4.4.1, where we develop the general expression for calculating the output sequence of certain types of discrete-time systems for an arbitrary input sequence. Program 2.3.m Program 2 4.m7 Chapter 2: Discrete-Time Signals in the Time Domain Since the unit step sequence andthe unit sample sequence are simply related through Fg. (2.47b), itis also possible to represent an arbitrary sequence asa weighted combination of delayed and advanced unit step sequences by making us of ths elation (Problem 2 44) 2.5 The Sampling Process WWe indicated earlier hat oftenthe discrete time sequence is developed by uniformly sampling a continuous time signal x(t), as illustrated in Figure ‘where the time variable ofthe c signal only at discret The relation between the two signals is given by Eq. (2. tinvous-time si ne instants, given by ‘ime lal is related to the time variable n of the discre =a 61) with Fr = 1/T denoting the san ling frequency and £ ‘or example, ifthe continuous-time signal is = 2xFr denoting the sampling angular frequency, F Rall) = Acosta fot + $) = Acos(Qat + 6), est the comespoding dsece-tme signal is given by fn] = Acos(QsnT +9) (2.63) where (2.64) isthe normalized angular fre frequency a is radians Signal is radians per sec Physical dimension of EXAMPLE 245 3 Hz, 7 He, rs jwency of the discret Per sample, while the ond and the unit of time, Aunt Signal x{n). The unit of the normalized angular Mt of the angular frequency 2, of the continuous-time the frequency J, is Herz ifthe sampling period has the Ambiguity inthe Dinero ences generated by uni and 13 Ha, respectively: gy (1) Me-Time Representation of Continuous-Time Signals srry gambling the tree cosine functions of frequencies F alOtt/see), ga(t) = con(l4irt/sec), and 2 Os NEMA GW Tm 01th dred soe ‘Plots of these sequences (shown with circle Chor steric ane Ph nye em a ewe 28 to associate a unique continuous reforms of frequencies given by pence sila} With being all cosine sampled at a 10-H2 rate ‘7 MnmeER meg. do tes2.5. The Sampling Process. 73 int ik] te! Tine signals. 2 (0) is shown with the solid Figure 2.28: Ambiguity in the discrete-time representation of continuous line, ¢2(0) is shown with the dasbed line, x(f) is shown with the dashed-dot line, and the se sampling is shown with circles. sence obtained by In the general case, the family of continuous-time sinusoids Kaalt) = Acos(E(Qot +6) +kQrt), k= 0, 41,42 leads to identi 1 sampled signals: edlT) = Acos (2p + kip)nT +4) = Acos (2#2e + kErIn +9) 2r 22m) Hi a tos (7 +4) A cos(aon +4) = x|n} 2.66) The above phenomenon ofa continuous-time sinusoidal signal of higher frequency acquiring the iden tty ofa sinusoidal sequence of lower frequency after sampling is called aliasing. Since there are an in- finite number of continuous-time functions that can lead to a given sequence when sampled periodically, aulditional conditions need tobe imposed so that the sequence (x{n}) = {x(n} ean uniquely represent the parent continuous-time function x (1). In which case, xq(1) can be fully recovered from a knowledge of {xf EXAMPLE 2.16 Mlustration of Allasing Determine the discrete-time signal r{n] obtained by uniformly sampling a continuous-time signal y(t) composed of a weighted sum of five sinusoidal signals of frequencies 30 Hz, 150 Hz, 170 Hz, 280 Hz, and 330 Ha, ata sampling rate of 200 He, as given below al!) = 6 c0s(607 1/sec)-+3 sin(30On /sec)+2 cos(340: 1 /see)-+4 c0s(S00m/sec) + 10 sin( 60% t/sec). The sampling period T = 1/200 = 0.005 sec. Hence, the generated discrete-time signal v[] is given by vf} = 6c0s(0-3ren) + 3sin(L Son) + 2c0s(1.720) + 4.cos(2.Sen) + 1sin(3.3xn) = 6cos(0.3n) + 3sin (2x —0.Sx)n) + 2c0s (2x —0.3n)n) + 4e0s (2 + 0.5m)n) + 10sin (4 ~0.7x)n) = 6 cos(03%n) — 3sin(0.S2n) + 2eos(0.3nn) +4 605(0.50) = 10sin(0.7)Aliasing Demo Chapter 2: Discrete-Time Signals in the Time Domain As canbe seen from the above, the components 3 sin(1Sn),.2. have been alased imo the components ~3 sin(0.5x1),2. 60s(1.72n), 4cos(2.Sn), and 10sin(3.3x) 605(0.3x7),4c0s(0.52n),and ~ 10sin(0.770), resuling ina discrete time sequence = Scus(0.3xn) + Se0s(0.5zn 40.6135) — 10 in(0.7xn), ‘composed of oly three simuscdal sequences of normalized angular frequencies: 0.3 I shuld be noted that an identical discrete-time signal is aso genera 200-1 sampling rate the following contimooutme signal sinusoidal continaous tte signals of fre nd by uniformly sampling at composed of a weighted sum ot quencies 30 H2, SO He, and 70 Ha: hve cos( Gt) + Seos(l00nt + 0.6835) — 10 sin(140x1 Anether example of a continuous time signal penering the sume discrete-time sequence is Met = Zeeatnt/tt)+4 coe) +1051) Goon sot se) 43 sn 7OOst/se) Saint Sompeset ofa weighed sum of ve sinusoidal cmtinoous time signals of frequencies 30 He. $50 He, 130 He, 280 He, and 350 Ha, respectively, 1 follows from Eq. (2.64) that if Oy > Of the discrete-time signal xa] the range frequency will o- then the corresponding normalize impling the parent continuous-time si On the other hand, if 2, < 2, Y yin the range — «digital frequency wy signal xa(¢) will be in o» the normalized digital ylnlsin— 6) = >» ylm + Ox{m) = rsy[-€). (2.68) ‘Thus, ryx(¢] is obtained by time-reversing the sequence rxy[C]. The autocorrelation sequence of x{n] is given by rel = > xladxin-¢ (2.69) obtained by setting y(n] = x(n] in Eq. 2.67). Note from Bg. (2.69) that 74.0 co X*ln] = &,, the energy of the signal x(n]. From Eq. (2.68), it follows that r,4{€] = rex{—€),implying that re,(€) is an even sequence fora real sequence xn} ‘An examination of Eq. (2.67) reveals thatthe expression for the cross-correlation looks quite similar to that of the convolution given by Eq. (2.204). This similarity is much clearer if we rewrite Eq. (2.67) as tel = YO xhnlyl-@ -m)] = x10 @ y-4. (2.70) It thus follows that the eross-correlation of the sequence x[n] with the reference sequence yn] is sim- ply given by the convolution sum of x{1] with y{-n], the time-reversed version of yi]. Likewise, the utocorrelation of x] is given by the convolution sum of itself with its time-reversed version 2.6.2 Properties of Autocorrelation and Cross-Correlation Sequences We next derive some basic properties of the autocorrelation and cross-correlation sequences {Pr096]. Consider two finite-energy sequences {2} and y{]. Now, the energy of the combined sequence a.x{n] +{n ~ ¢ivalso finite and non E ceil t y—e + 2arsy{e) + r,5[0] > 0, where rss[0] = 8 > 0 and ryy[0] = 6, are energies of the sequences We can rewrite Eq te felt ola) for any finite value of a. Orin other words. the matrix [rt so] eelOlry (0) — is postive semi-definite, Ths implies oF equivalently, MraylQl quality provides an uy HPPet bound for the cross-corrclation se: X{n], the above reduces to quence samples. If we set 2B) reall) < rexlO] = By @ This is significant result sequence has it states tha at 7er0 lag ( = 0), the Toderive an additional property of the Sample value of the autocorrelation ‘“toss-conrelation sequence, consider the case + brn Nj, where Vis an in raed} > Ois an arbitrary number. la his case &, = 6, ,and therefore, VEE; = [8 = be, Using the above rest in Eg, (272) brss10) NV. The autocorrelation of wf} is given by ae rovlll= 47 DL whluh—4 Cin] + din) (tn - 9 + din- mer d{n| din — 4) 1 -4+5 Mo iS 1 + 7p Hida ap dis a reel€] + raall] + reall) + resid. (2.80) Now in the above equation, rz3{(] is a periodic sequence with a period NV and hence, it will have Peaks at £ = 0,N, + with the same amplitudes as € approaches M.. As (n] and d{n) are not ‘correlated, samples of cross-correlation sequences rq (6) and raz) are likely to be very small relative to the amplitudes of r,.(4]. The autocorrelation of the disturbance signal d{n] shows a peak at € = 0, with Cother samples having rapidly decreasing amplitudes with increasing values of |f|. Hence, the peaks of row] for ¢ > Oare essentially due to the peaks of rex{2] and can be used to determine whether Z[a] isa Periodic sequence and its period if the peaks occur at periodic intervals. 2.6.6 Correlation Computation of a Periodic Sequence Using MATLAB Using MATLAB we determine the period of a noise-corrupted periodic sequence in Example 2.19,Program 2.6.m Chapter 2: Discrete-Time Signals in the Time Domain neh? 6 nae L I ® 4 J oe Pengte caf ot PT tg fing | i eth call og Tt rer Lag index." , Lag index ni Figure 231: (a) Autocorrelation sequence of ‘he nose-corrpted sinusoid and (b) autocorrelation sequence of the EXAMPLE219 Determination ofthe Period of a ‘We determine the period additive weiformly distribu Noise-Corrupted Periodic Sequence the sinaoidal sequence x)a] = cos(0.25n),0-< n 95, corrupted by an tstrandom noise of amplitude in the range [-0.5,0 5 Soidal sequence, ig Hn. theres a tong peak a em ag, However, thes een ee inbcting the pei the sino sequen pp shows the plo ofthe amocorration sequence rg this plo rae) shows avery rong pe other values ofthe lag of the avise-corrupted simu Figure 231(a), As iatnct peaks at lags that as expected.* Figur lf the noise component. As can be seen from =X oly at 27 lag. The amplitudes ate considerably salle a ‘sample values ofthe noise sequcnce are uncarelated wt 2.7 Random Signals Most of the book deals with sumption of this type o processe he processin if discrete-time si such a8 a mathematical expression or Aicterministc signal since all sample value inet For example the cui eases deterministic Sequences, 4 Si ig Of signals that are eterministc in n als is th at they can be uniquely det 4 rule or & lookup table. Such of the sequence are well defined of Eq. (2.48) and the exponential ature “termined by well-defined 8 signal is usually called The underlying for all values of the time Comprise another class cannot be reproduce be modeled using st are models for 5 between the Practical analog jot be predicted abead oF a stochastic signal, al, and therefore needs to ‘amples of random signals ignal generated by forming the diff Severating the sign the signal. Some signals. The error si atstical information about Pech, music, and seismi common ex: ideal samp bya al for analysis purposes.” The red version generate Se Sec lites the pea NE eto i them fr aang les fhe ag ‘14x 1 ae dae to the 1 pred i the nite lengths of the perioic cOMpation ofthe an2.8. Summary 81 noise sequence d (n} of Figure 2.6(b) generated using the rand function of MATLAB is also an example ofa random signal. ‘The discrete-time random signal or process consists of a typically infinite collection or ensemble of discrete-time sequences {X ]}. One particular sequence in this collection {x{n}} is called a realization of the random process. At given time index n, the observed sample value {1} is the value taken by the random variable X(n). Thus, random process is a family of random variables {X n]}. In general, the ange of sample values is a continuum. We review in Appendix C the important statistical properties of the random variable and the random process 2.8 Summary In this chapter, we introduced some import tal concepts regarding the characterization of discrete-time signals in the time domain, Certain basic discrete-time signals that play important roles in discrete-time signal processing have been defined, along with various mathematical operations used for generating more complex signals. The relation between a continuous-time signal and the discrete- time signal generated by sampling the former at uniform time intervals has been examined. Finally, the concepts of the autocorrelation of a sequence and the cross-correlation between a pair of sequences are introduced, Further insights can often be obtained by considering the frequency-domain representations of discrete- tifne signals discussed in Chapter 3. We consider the time-domain representation of discrete-time systems and the frequency-domain representation of a certain class of discrete-time systems in Chapter 4 2.9 Problems 2.1 Determine the £1-, £2-,and L.y-norms ofthe following finite-length sequences: () (eyfal} = (121, 342, 10.01, -0.13, -5.1, -129, 3.87, 246, -3.21, a0, ©) (rafal) = (981, 622, 1.17, O81, 0.12, 621, =12.01, 3.16, ~0.14, 1.87) 22 Show that fx allt 23 Consider the following sequences: xh]=2, 0, -1, 6 -3. 2 0) -3<0<3, The sample values of each ofthe above sequences outside the ranges specified are all eros. Generate the follow ing sequences: a) en} = fn +3], 0) df © ela @) ula) = xin =3} + yin +3) (©) vba] = yfn ~ 31-wln-+ 21, © sin] = yf +4] w[n—3}.and ©) rn] = 3.90).82 Chapter 2: Discrete-Time Signals in the Time Domain 24 Figure F2.1 shows the schematic representation of four operations developed using the three basic operations ‘ution, muiipication and delaying. Develop the expression for ylni for cach operation as a func 7 Wl—w> 949 +n © tat Baie . Aa ir sin St comolton of a enti M seqpece wth « Hengit-W sequence leads to a sequence of lengih 26 Letlal.yinl.and win] denote thee inion ‘ample of each sequence occuring at po eth Sequences of lengths NM. and L.. respectively, with thefts - Whats the length ofthe segue ace x[a] ® yln] ® w(n]? Conse the flowing caus fit dengh QUENCES With their fis samples at n — 0: 218 1G 1 1 1 Som that 0 sl} ® 22f0] = x5hn]@ safe) 218 Bvaluate the tin a nea Olt lining rsmencs with iat Ap st“ eee, Maen 0 pect Hhl=t-1, 2, 9, 1 29 Determine the f tained by a inca convolatio of the sequences given in Problem 2:3: ©) onl = fo} @ win i ib © tel = win} @ y/o 210 Let MU S2h— 1] Bere on a een of yn2.9. Problems 83 2 ee sl] = x1 fe] © x2fn] ® x3fa] and Aa) = xy Jn — Na] @ x2{n — a} @ xa\n — Ns]. Express h(n] in terms of st 242 Consider the following finit i) thin}, —M R. Define (a) vuln] = Ale] © An, (©) y2lnl = ge] © fn) (©) ysin] = wha @ win, (®) vain] = Alen] © sl (©) y(n] = Alm] @ win} ‘What isthe length of each of the convolved sequences? What isthe range of the index m for which each of the ‘above convolved sequences is ‘ined’? 2.13 Let y{n] denote the linear convolution of the two sequences {al} = (2, —3, 4. hln}} = [-3, 5, ~6, 4.) —2 0, whereas an be recovered fram ts ad pat xan] only fra (b) Is it pons wo fully re rer A causal complen sequence s(n}? Jusity your answers onjugate symmetric pty ramble secpence. me 10 By. (2.4) is a periotic sequence with a Show thatthe seque peed 8 Nery he elation between the uit sample sequence I+] snd the unit sep sequence j[n] given in Eq. (2470) Peermine the even and od pany fhe oing al sequences: ) aah i ©) sla 1 228 Which ones ofthe folowing sequences are bounded fa where A and a ace complex runters nd 1 © Ye] = own ~ 1), whee je} <1 in} = ( in —1}, tha he sequence xin A Hle ~ 1} is nt absoty ely summable even though 2.30 Show that he WERCES Ae AbLolately summabie: nce Fln] formed by an N periodic extension2.9, Problems 85 231 Show that the following sequences are absolutely summable. () xalel = Frulel. 0) x60] = eatery nln] 2.32 Show that an absolutely summable sequen surmable c has finite energy, but a finite energy sequence may not be absol 233 Show thatthe square-summable sequence x; 1] of Eq, (2.35) is not absolutely summable, 2.34 Show thatthe sequence x2{n] = ®°0" y(n — 1 is square-summable bi not absolutely summable, 2.35 Compute the energy ofthe following sequences: (@) xehh=Aa"abllal<1, —) xsinl= chute 1) 2.36 Determine the average poster and the energy of the following sequences: () ale] yt ©) sala) = nba, ©) xsin] = nln), (@) xaln] = Age", (©) sin] = Acos( 25 + 9), 2.37 Determine the samples of one period of the periodic sequences obtained by an N-petiodic extension of the sequences of Problem 23 forthe following values of N: (a) N = 6,and (b) N= 8. 2.38 The following sequences represent one period of a sinusoidal sequence of the fo @ Uk =. 1, (5, -05. 05, -05} (©) (0, 0.5878, -0.9511, 0.9511, 05878) @ 20 -2 0} Determine the values of the parameters A <%o,and ¢ for each case 229 Determine the fundamental period ofthe following periodic sequences: (2) al] = e/028e0, 0) ipl <0s(0.6rn +032), (©) Hela] = Re (e/*™/8) + 1m(e/#H/5) () Sgln] = 6sin(0.15x0) ~ cos(0.120 + 012), (©) Se(n] = sin(0.1un +0:75x) ~ Se0s(0.8xn + 0.22) + cos( 1.3) 240 Determine the fundamental period of the sinusoidal sequence £] = A in(agn) forthe following values of the angular frequency ey @ 03m, O48x, (0) OAS, O2SM. (O75. OTS. 2A Determine the period of the sinusoidal sequence [1] = sin(0.06-rn), Determine at last tw Sinusoidal sequences having the same period 285]. other distinet86 Chapter 2: Discrete-Time Signals in the Time Domain 242 Deterine the fundamental period andthe average power ofthe following periodic sequenc (0) aha) = (©) ishn] = 2eos(2nn (1) R4ln] = 3cos(Sn/7) (©) isla] = &cos(2"n/5) + 30os(320/5), ) Sohn] = 4cos(sun/3) 2AY Express the sequences Pil ofroses!2.3 asa linear combination of delayed unit sample sequence lel, yl} and win] of Problem 23 #1 linear combination of delayed unit sep sequences 24 Exes te length sequence sa] = (21, 34, <1. 49) _) sequence jn 24S: Bxpess the the sequence 4 terms ofthe unit step sequcrce jf 24 Deep cond for exesions ete fowing convolution suns ale) © Ho), 0 mera} @ pla 247 Consider te fciowing sequences: above sequences: se) stn — a9 hn + M)~ nin — 9 where Vand Mare posi Determine the lcaion an he tt Y > crcl Kaen ant the value of wT Mse ofthe flowing sequences witout performing the2.10. MATLAB Exercises 87 (2) vila] = x(n) @ xf}, (©) yalnl = hin @ hi}, © yslo] = gla] @ gla}, yaln] = slo @ elo} © xin] @ hla, xn] ® ol (&) yolm] = x(n] @ wf) 249 Consider two complex-valued sequences hi] and ¢n] expressed as a sum of their respective conjugate sym. 1 = ses)"] + geal). For each of the following sequences, determine if itis conjugate symmetric or conjugate antisymmetric metric and conjugate antisymmetric pats, ic. hin] = hen] + heal, a @) hale] @ xclnl, —(b) heal] @ eslul, (©) Neale] @ gealn) ous-time signal x(t) = Acos(Qot + $) can be uniquely recovered from its sampled 2x/T > 28g, 2.50 Show that the cont version x{n] = x9(nT), 20 < n < 50, if the sampling frequency Qr 251 A continuous-time sinusoidal signal xq(¢) = cos ot is sampled at t = nT’, ~c0

You might also like

• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (821)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (122)
• • Magazines
  • Podcasts
  • Sheet music
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4610)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1840)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Tóibín

  Rating: 3.5 out of 5 stars

  3.5/5 (1946)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (74)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2259)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1104)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (897)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2520)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4203)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (537)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (345)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (792)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2104)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (589)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1016)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1898)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (842)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (440)
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (599)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (806)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (401)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1715)
• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5806)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brene Brown

  Rating: 4 out of 5 stars

  4/5 (1091)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)
• Buku Istilah Kejuruteraan

  

  Document765 pages

  Buku Istilah Kejuruteraan

  W Mohd Zailimi Abdullah

  89% (18)
• Doa Abu Darda

  

  Document1 page

  Doa Abu Darda

  eddyonn

  No ratings yet
• Doa Diamalkam Ketika Rukuk Dan Sujud

  

  Document1 page

  Doa Diamalkam Ketika Rukuk Dan Sujud

  eddyonn

  No ratings yet
• Intipati Al Quran

  

  Document30 pages

  Intipati Al Quran

  eddyonn

  No ratings yet
• Lelaki Dayus

  

  Document4 pages

  Lelaki Dayus

  eddyonn

  No ratings yet
• Cara Buat Baja Organik Cecair Supaya Buah Besar

  

  Document2 pages

  Cara Buat Baja Organik Cecair Supaya Buah Besar

  eddyonn

  No ratings yet
• Amalan Waktu Malam

  

  Document16 pages

  Amalan Waktu Malam

  eddyonn

  No ratings yet
• Booster Liguano X70 Berbuah Diluar Musim

  

  Document26 pages

  Booster Liguano X70 Berbuah Diluar Musim

  eddyonn

  No ratings yet
• Koleksi Doa Dan Zikir PDF

  

  Document100 pages

  Koleksi Doa Dan Zikir PDF

  eddyonn

  No ratings yet
• 60 Jenis Tanaman Obat

  

  Document61 pages

  60 Jenis Tanaman Obat

  eddyonn

  No ratings yet
• Ayam Tandori

  

  Document2 pages

  Ayam Tandori

  eddyonn

  No ratings yet
• Memasukkan Simbol Saintifik Dalam MS Word

  

  Document6 pages

  Memasukkan Simbol Saintifik Dalam MS Word

  eddyonn

  No ratings yet
• Rahsia Surah Yassin Ayat Ke 58

  

  Document1 page

  Rahsia Surah Yassin Ayat Ke 58

  eddyonn

  No ratings yet