950--951 for discrete-time signals Accumulator 55 with normalized algorithm 829-830 Adaptive arrays 900-902 949-950 moments 825 Adaptive filters 880--959 Adaptive line enhancer power 826 applications of 880--902 895-896 time-averages 830--833 for antenna arrays Adaptive noise cancelling 900--902 896-897 B for channel equalization Akaike information criterion 883-887 (AIC) 997 Backward predictor 578, 841 for echo cancellation Algorithms 4 Bandlimited signals 266 887-891 Chirp-z 544-549 Bandpass filter 327, 332-334 for interference suppres- FFT 511-537 Bandpass signal 266 sion 891-895 Goertzel 542-544 Bandwidth 265-267,660 for linear predictive Remez 686-691 Bartlett's method (see Power coding 897-900 Aliasing, frequency-domain spectrum estimation) for noise cancellation 20, 389-394 Bessel filter 726-727 896-897 time-domain 391 Bilinear transformaiion for system identification Alternation theorem 684 712-717 882-883 Amplitude 14 Binary codes 405 for system modeling Analog signals (see Signals) Blackman-Tukey method 880-882 (see Power spectrum Analog-to-digital (AID) direct-form FIR 902-927 estimation) converter 5,19, properties of 925-927 Burg algorithm (see Power 401-410 with FAEST algorithm spectrum estimation) Autocorrelation, of determinis- 946 Butterworth filters 717-720 tic signals 116-128 with fast RLS algorithm of random signals 323-326, 923-925, 945-947 826 c with MSE algorithm Autocovariance 826 Canonic form 112 903-907 Autoregressive (AR) process Capon method (see Power with RLS algorithm 836,987 spectrum estimation) 907-927 autocorrelation of 837 Cauchy integral theorem with square-root (LDU) Autoregressive-moving 156 algorithms 921-923 average (ARMA) Causality, implications of lattice-ladder filters 927-954 process 837, 987 65~59 properties of 951-954 autocorrelation of 837 Causal signals 85 with a priori LS algorithm Averages, autocorrelation Causal systems 66-67, 83-85, 940 826 196-198 with error-feedback autocovariance 826 Cepstral coefficients 835 algorithm 944 ensemble 825-828 Cepstrum 261-262 1077 1078 Index
99 98-101 periodicity 465 Chebyshev filters 720-724 particular solution 101-103 symmetry 465-470 Chirp signal 547 total solution 103-108 table 470 Chirp-z transform algorithm Differentiator 691 time reversal 476 544-549 design of 691-693 relationship to Fourier series Circular convolution 470-476 Digital resonator 335 461,463 Coding 20-35 Digital sinusoidal oscillator relationship to Fourier Comb filter 341-344 347-349 transform 461 Conjugate-gradient algorithm Digital-to-analog (D!A) relationship to z-transform 906 converter 5, 20, 36, 462 Constant-coefficient difference 408-440 use in frequency analysis equations 93-108 Dirichlet conditions, for 488-495 solution of 98-108 Fourier series 228 use in linear filtering Continuous-time signals 17 for Fourier transform 237 480-488 exponentials 17 Discrete Fourier transform Discrete-time signals 9, 36-52 sampling of 17-22, 384-394 (DFf) 454-461 antisymmetric (odd) 48 sampling theorem for computation 480-488, correlation 116-128 26-31, 384-394 511-536 definition 9, 42 Convolution (linear) 69-80 butterfly 522, 526, 528 exponential 44-45 circular 470-476 decimination -in -freq uency frequency analysis of properties 80-83 FFf algorithm 524-526 241-259 sum 69 decimination-in-time FFf non periodic 48 Correlation 116-128,827 algorithm 519-526 periodic 11-17 autocorrelation 120, direct 480-488,511-513 random 11 321-326, 826 divide-and-conquer representation of 36 computation 123-125 method 513-536 sinusoidal 14-16 cross-correlation 118, in-place computations symmetric (even) 48 321-323 523 unit ramp 44 of periodic signals 123 radix-2 FFf algorithms unit sample 43 properties 120-123 519-526 unit step 43 Coupled-form oscillator nrdix-4 FFf algorithms Discrete-time systems 53-69 347-349 526-532 causal 66-67, 83-85 Cross-power density spectrum shuffling of data 523 dynamic 60 829 split radix 532-536 finite-duration impulse via linear filtering response 88-89 537-549 finite memory 60, 88-89 D definition 456 implementation of 563-601 Dead band 625 IDFf 456 infinite-duration impulse Decimation 754-760 implementation of FFT response 88-89 Deconvolution 262, 349-354, algorithm 536-539 infinite memory 60, 88-89 358-360 properties 464-480 linear 63 homomorphic 262, 360-362 circular convolution memory less 57 Delta modulation 434 470-476 noncausal 66-67 Difference equations 89-108 circular correlation 478 nonlinear 64 constant coefficient 98-108 circular frequency shift nonrecursive 92-93 solution 98-108 478 recursive 90-93 for recursive systems 93, circular time shift 477 relaxed 56 108 complex conjugate 478 representation 36 from one-sided z -transform linearity 465 shift-invariant 60-61 210-211 multiplication 470-476 stability triangle 202 Index 1079
Discrete-time systems (contin- F prediction error filter 575,
ued) 839 Fast Fourier transform (FFT) stable (BIBO) 67, 85-88 smoothing 36 algorithms 511-537 static 60 structures· 563-582 application to 511 time-invariant 60-61 Wiener filter 862-873 application to, correlation unit sample (impulse) 537-539 Filter banks 790-796 response 73-80 critically sampled 794 application to, efficient unstable 67,85-88 computation of DFT quadrature mirror 798-799 Distortion, amplitude 511-537 uniform DFT 790-796 312 delay 328 application to, linear Filtering 480 harmonic 366 filtering 537-539 of long data sequences phase 312 485-488 implementation 536-537 Down sampling 52 overlap-add method for mirror FFT 536 (see also Sampling rate 487-488 phase FFT 536 conversion) 52 overlap-save method for radix-2 algorithm 519-526 Dynamic range 33, 404, 605 485-487 decimation-in-frequency via DFT 481-488 524-526 Filter transformations E decimation-in-time 334-338, 727-734 519-526 Echo, far-end analog domain 730-732 radix-4 algorithm 526-532 digital domain 732-734 888 nea~end 888 split-radix 532-536 Echo canceller 888 lowpass-to-highpass 334-338 Fast Kalman algorithms 945 Final prediction error (FPE) Echo suppressor 888 Fibonacci sequence 210 Eigenfunction 302 criterion 996-997 difference equation 210-211 Final value theorem 209 Eigenvalue 302 Filter 326 Elliptic filters 724-726 FIR filters 660 bandpass 327, 332-334 anti symmetric 660-670 Energy definition 312, 326-329 design 660-701 definition 45 design of IIR filters comparison of methods density spectrum 238-241, 326-349, 701-727 699-701 254-259 all pass 345-347 differentiators 691-693 partial 382 by pole-zero placement equiripple (Chebyshev) signal 45-46 329-349 approximation 678-699 Energy density spectrum comb 341-344 frequency sampling 238-241, 254-259 notch 338-341 method 671-678 computation 961-966 resonators (digital) Hilbert transformers Ensemble 825 335-338 693-699 averages 825-828 design of linear-phase FIR window method 678-691 Envelope delay 328 660-701 linear phase property Ergodic 830 transition coefficient for 664-670 correlation-ergodic 832-833 1047-1052 symmetric 660-670 mean-ergodic 831-832 distortion 312 FIR filter structures 565-582 Estimate (properties) distortionless 328 cascade form 567-568 asymptotic bias 969 frequency-selective 326 direct fQrm 566-567 asymptotic variance 968 highpass 327,329-332 conversion to lattice form bias 969 ideal 327-329 581-582 consistent 968 lowpass 327, 329-332 frequency sampling form (see also Power spectrum nonideal, passband ripple 569-574 estimation) 659 stopband ripple lattice form 574-605, 859 Estimate (properties), variance 660 conversion to direct form 968 transition band 660 579-581 1080 Index
FIR filter structures (contin- highest 16
ued) negative 13 normalized 22 IIR filters 701-727 transposed form 586-588 positive 13 design from analog filters FIR systems 88, 92, 113 701-727 Fixed-point representation relative 22 Frequency analysis 234-251 by approximation of 601-605 derivatives 703-707 Floating-point representation continuous-time aperiodic signals 234-238 by bilinear transformation 605-608 712-717,727 Flowgraphs 584-586 continuous-time periodic signals 226-233 by impulse invariance Folding frequency 25,389-391 707-712 Forced response 95 discrete-time aperiodic signals 248-251 by matched-z transforma- Forward predictor 578, tion 717 838-841 discrete-time periodic signals 241-245 least-squares design Fourier series 18, 226-233, methods 746-747 241-245 dualities 267-268 for LTI systems 300-326 Pade approximaf ~n coefficients of 226-229, table of formulas for 269 747-748 241-245 Frequency response 306 pole-zero placement for continuous-time periodic computation 317-321 329-349 signals 226-233 geometric interpretation of Prony's (least squares) for discrete-time periodic 317-321 746-747 signals 241-245 magnitude of 304 Shanks' (least squares) Fourier transform 234-238, phase of 304 748-749 248-251 relation to system function IIR filter structures 582-601 convergence of 251-254 inverse 236 314-317 cascade form 588-591 of continuous-time aperiodic to exponentials 301-309 direct form 582-584 signals 234-238 to sinusoids 306-309 lattice-ladder 594-601, of discrete-time aperiodic Frequency transformations (see 860-862 signals 248-251 Filter transformations) parallel form 591-594 properties 283-291 Full duplex transmission 887 second-order modules 589 convolution 283-284 Fundamental period 15 transposed forms 584-588 correlation 284 Impulse response 106-108 G Initial value theorems 168 differentiation 289-291 frequency shifting 286 Gaussian random variable Innovations process 833-836 linearity 279-281 1045-1046 Interpolation 28-31, 389, 751, modulation 286-287 subroutine for 1045 760-762 multiplication 288-289 Gibbs phenomenon 254, 669 function 28 of signals with poles on Goertzel algorithm 542-544 ideal 28-31, 389 unit circle 262-265 Granular noise 407 linear 36,427-440 Parseval's theorem 287 Group (envelope) delay 328 Inverse filter 349-350 relationship to Inverse Fourier transform z-transform 259-261 H 236, 251 symmetry 271-279 Harmonic distortion 366, Inverse system 350-351 table 290 446-447 Inverse z-transform 156-170, time-reversal 282 High-frequency signal 265 179-193 time-shifting 281 Hilbert transform 658 by contour integration Frequency 11-16 Hilbert transformer 693-699 156-157, 180-182 alias 15,23 Homomorphic 360, 362 integral formula 156 content 26 deconvolution 360-362 partial fraction expansion folding 25, 389-391 system 361 184-193 fundamental range 17 Hybrid 888-891 power series 182-184 Index 1081
J Linear prediction filter (see Natural response 95
Joint-process estimate 938--940 Linear prediction) Natural signals 267-268 575 Noise subspace 1017 K Linear predictive coding 897 Noise whitening filter 835 of speech 897-900 Normal equations 846 Kalman gain vector 918 LMS algorithm 905-907 solution of 846--854 L excess mean-square error of Levinson-Durbin algo- 908 rithm 846-850 Lattice filter algorithms properties of 907 Schur algorithm 850-853 927-954 Local loop 888 Number representation a posteriori form 940 Low-frequency signal 265 601-608 a priori form 940 Lowpass filter 327 fixed-point 601--605 error-feedback form 944 LTI systems 112-116 floating point 605-608 gradient form 950-951 moving average 113 Nyquist rate 28 joint process estimate second order 112-113 938--940 modified form 940-946 structures 108--116 canonic form 112 o normalized form 949-950 direct form I 108--109 One's complement. 603 properties of 951-954 direct form II 109-113 One-sided z-transform square-foot form 949-950 nonrecursive 112-116 205-211 Lattice filters 574--579 recursive 112-116 Orthogonality principle 594--601, 858--862, 874 weighted moving average 866-867 ARMA structure 860-862 112 Oscillators (sinusoidal genera- AR structures 858--860 tors) 347 MA structure 838--841 M CORDIC algorithm for LDU decomposition 921-923 349 Leakage 489, 964 Maximal ripple filters 685 coupled-form 347-349 Learning curves 911 Maximum entropy method digital 347 Least squares 746-747 993 Overflow 629--631 filter design 746-747 Maximum-phase system 354--357 Overlap-add method 487-488 Least-squares estimation Overlap-save method 485-487. 907 Mean square estimation 866 903-905 ' Overload noise 407 Levinson-Durbin algorithm Oversampling NO 433-440 846-850 orthogonality principle 866-867 Oversampling D/A 439 generalized 850, 876 Minimum description length split Levinson 874 p (MDL) 997 Limit cycle oscillations Minimum-phase system Paley-Wiener theorem 656 624--629 354-357 Parseval's relations 238 255 Linear filtering 480-488 Minimum variance estimate 287, 479 " based on DFf 480-488 1012-1015 aperiodic (energy) signals overlap-add method Mixed-phase system 354--357 238, 255, 287 487-488 Moving-average filter 304 DFf 479 overlap-save method Moving-average (MA) process periodic (power) signals 485-487 837,987 230,246 Linear interpolation 427-440 autocorrelation of 837 Partial energy 358 Linear prediction 578, Moving-average signal 115 Partial fraction expansion (see 838--858 Multichannel signal 8 Inverse z-transform) backward 841-845 Multidimensional signal 6-9 Periodogram 966-971 forward 838-841 lattice filter for 845 estimation of 966-971 N normal equations for 846 mean value 968 properties of 855-858 Narrowband signal 266 variance 968 /' 1082 Index
Phase 12 computational require- of WeIch estimate 982-983
maximum 354-357 ments 984-985 Quantization 20, 31-35, minimum 354-357 performance characteris- 403-406 mixed 354-357 tics 981-984 differential 433 reSponse 306 WeIch 975-977,982-983 differential predictive 434 Pisarenko method 1015-1019 parametric (model-based) dynamic range 33, 404, 605 Poles 170 methods 985-1009 error 31, 34, 613-639 complex conjugate 189-192, ARMA model 988, in AID conversion 403-406 204-205 999-1001 in filter coefficients 613-620 AR model 989 level 32, 403 distinct 173-198 AR model order selection resolution 32, 605 location 173-198 996-997 rounding 32, 608-612 multiple-order 187-188 Burg method 990-994 step size 32, 605 Polyphase filters 766-767 least-squares 994-995 truncation 32, 608-612 for decimation 768 MA model 990, 997-999 Quantization effects 31, 405, for interpolation 773 maximum entropy method 549, 601 Power 46 993 fixed-point numbers definition 46 model parameters 601-605 signal 47 988-990 one's complement 603 Power density spectrum modified Burg 991 sign-magnitUde 603 229-233 relation to linear predic- table of bipolar codes definition 230 tion 988-990 406 estimation of (see also Power sequential least squares two's complement spectrum estimation) 995-996 603-605 230 Yule-Walker 990 floating-point numbers periodic signals 229-233, USe of D FT 971-973 605-608 245-248 Prediction coefficients 839 in AID conversion 34-35, random signals 828-830 Prediction-error filter 574, 839 406-408 rectangular pulse train properties of 855-858 in computation of DFT 232-233 Principal eigenvalues 1020 549-555 POwer spectrum estimation Probability density function direct computation 963 824-825 549-551 Capon (minimum variance) Probability distribution FFf algorithms 551-555 method 1012-1015 function 1041-1044 in filter coefficients 613-620 direct method 963 Prony's method 746-747 limit cycles 624-629 eigenanalysis algorithms Pseudorandom sequences dead band 625 1019-1028 145 overflow 629-631 ESPRIT 1'022-1025 Barker sequence 145 zero-input 625 MUSIC 1021 maximal-length shift register scaling to prevent overflow order selection 1025-1026 sequences 144-145 629-631 Pisarenko 1015-1019 statistical characterization experimental results Q 631-639 1001-1009 Quantizer 403 Quadrature mirror filters from finite data 966-973 midrise 404 788 indirect method 963 midtread 404 for perfect reconstruction leakage 964 resolution 403-405 798-799 nonparametric methods uniform 404 for subband coding 788 973-985 Quality 981-984 Bartlett 974-975 of Bartlett estimate 982 R Blackman-Thkey of Blackman-Tukey estimate Random number generators 977-981, 983-984 983-984 1041-1046 Index 1083
Random processes 323-326, Reflection coefficients 575, of bandpass signals 779-781
824-833 598,845-846,927 polyphase filters for averages 825-831 Resonator (see Digital 766-767 autocorrelation 826 resonator) Sampling theorem 26-28, autocovariance 826 Reverse (reciprocal) polyno- 384-394 expected value 825 mial 579, 842 Schur algorithm 850-853 for discrete-time signals backward system function pipelined architecture for 829-830 579,842 853-854 moments 825 Round-off error 608-612, split-Schur algorithm 874 power 826 631-639 Shanks'method 748-749 correlation-ergodic 832-833 Sigma-delta modulation 436 discrete-time 829-830 s Signal flowgraphs 584-588 ergodic 830 Sample-and-hold 402-403, 409 Signals 2-4 jointly stationary 825 Sample function 825 analog 9 mean-ergodic 831-832 Sampling 9, 19, 21, 384-394 antisymmetric 48 power density spectrum aliasing effects 25-26 aperiodic 48 828-829 frequency 21 bandpass 266 response of linear systems frequency domain 449-454 continuous-time 9 323-326 interval 21 deterministic 11 autocorrelation 323-326 Nyquist rate 28 digital 10 expected value 323 of analog signals 21-31, discrete-time 9, 36-52 power density spectrum 384-394 electrocardiogram (EeG) 324-326 of discrete-time signals 8 sample function 825 751-806 harmonically related 17 stationary 825 of sinusoidal signals 22-26 multichannel 8 wide-sense 826 period 21 time-averages 830-832 multidimensional 9 periodic 21 natural 267 Random signals (see Random rate 21 processes) frequency ranges 267-268 theorem 26-28 Rational z-transforms 184-193 periodic 13 time-domain 22-26,384-394 poles 170-173 random 11, 824-833 uniform 21 zeros 170-173 correlation-ergodic Sampling-rate conversion Recursive least squares 832-833 762-806 907-954 ergodic 824 (see also Sampling-rate direct-form FIR algorithms conversion) expected value of 825 907-927 applications of 784-806 mean-ergodic 832-833 fast LS 923-925,945-947 for DFf filter banks moments of 826-830 properties of 925-927 790-796 statistically independent lattice algorithms 928-954 for interfacing 785 827 a posteriori form 940 for lowpass filters 786 strict-sense stationary a priori form 940 for phase shifters 784-785 825 error-feedback form for subband coding time-averages 830-833 944 787-788 unbiased 831 gradient 950-951 for transmultiplexing uncorrelated 827 joint process estimate 796-798 wide-sense stationary 938-940 by arbitrary factor 781-782 826 modified form 940-946 by rational factor 762-766 seismic 268 normalized form 949-950 decimation 751-760 sinusoidal 12 properties of 951-954 filter design for 762-775 speech 2-4 square-root form 921-923 interpolation 751,760-762 symmetric 48 Recursive systems 112-116 multistage 775-779 Signal subspace 1020 1084 Index
Sign magnitude representation steady-state 195-196 y
603 transient 104, 195-196 Sinusoidal generators (see zero-input 95 Yule-Walker equations 846 Oscillators) zero-state 94 modified 1000 Spectrum 225-226 Yule-Walker method 990 analysis 226 T estimation of 226,961-1028 (see also Power spectrum Time averages 830-833 z Time-limited signals 266 estimation) 226 Zero-input linear 96 Toeplitz matrix 847,864 Split-radix algorithms 532-536 Zero-input response 95 Transient response 104, Spread-spectrum signal 892 195-196, 309-311 Zero-order hold 34, 409 Stability of LTI systems Transition band 660 Zero padding 456 196-203 1Tansposed structures 584-588 Zeros 170 of second-order systems Truncation error 32, 608-612 201-203 Zero-state linear 96 Trunk lines 888 Zoom frequency analysis Stability triangle 202 Two's complement representa- 821-822 Steady-state response tion 603 195-196, 311-312 z-transforms 147 Structures 108-116 definition 147-148 direct form I 108-109 u bilateral (two-sided) direct form II 109-113 Uniform distribution 407, 147-148 Subband coding 787-789 549-551, 608-612 unilateral (one-sided) Superposition principle 63 Unit circle 261-265 205-211 Superposition summation Unit sample (impulse) inverse 156-170, 179-193 73 response 106-108 by contour integration System 3, 53-56 Unit sample sequence 43 156-157, 180-182 dynamic 60 by partial finite memory 60 v fraction-expansion infinite memory 60 184-193 Variability 981 inverse 350 by power series 182-184 Variance 549-551,631-639 invertible 350 properties 157-170 relaxed 56 convolution 164-166 System function 177-179, w correlation 166-167 314-317 Welch method 975-977, differentiation 163-164 of all-pole system 179 982-985 initial value theorem of all-zero system 177-179 Wide band signal 266 168 of LTI systems 177-179 Wiener filters 862-873, 904 linearity 157-159 relation to frequency FIR structure 864-866 multiplication 167-168 response 314-317 for filtering 863 Parseval's relation 168 System identification 350, for prediction 863 scaling 161-162 358-360 for smoothing 863 System modeling 836 IIR structure 867-871 table of 169 System responses 94 noncausal 871-873 time reversal 162 forced 95 Wiener-Hopf equation 864 time shifting 159-161 impulse 106-108 Wiener-Khintchine theorem rational 170-179 natural (free) 95,212 285 region of convergence of relaxed pole-zero systems Window functions 668 (ROC) 147-156 170-179 Wold representation 836 relationship of Fourier of systems with initial Wolfer sunspot numbers 10 transform 259-261 conditions 211-214 autocorrelation 125 table of 169