74 Random Vectors Chap. 2 PROBLEMS 2.1. A probability density function for a two-dimensional random vector is defined by { Adda 11y%2 > Oand xs +02 £1 foo) = 0 otherwise (a) What is the distribution function Fz(8)? Use this sesut to find the numerical value of the constant A. (b) What isthe marginal density f.,(00)? Let x and y be random variables (one-dimensional random vectors) with density functions {5 OP y and = ate not uncorrelated and 23 and 4 are not uncorrelated, ft] 2.24, Given the correlation marie andthe transformation%p.4 al to dom tion Problems a5 Probability, Random Processes, and Estimation Theory for Jewood Clifs, New Jersey, 1986. bent Probabilistic Models in Engineering Sciences, Volum U Henry Stark end John W. Woods. Engineers. Prentice Hall, [ne., Ena 15, Harold J. Larson and Bruno O. Schut John Wiley & Sons, New York, 1979. Kenneth §, Millet, Complex Stochastic Processes, Addison-Wesley, Reading, Massachuseis. 14 16. 1974 17; Jemyy M. Mendel. Turi on higher-order statistics (speci) in signal processing one sysemt faci theoretcal results and some applications. Proceedings ofthe IEEE, 79(3}278-305, March 1991 jgnal processing and system theory: a short 18, Jemy M. Mendel. Use of higher order statistics in si perspective, In Proceedings of 8° International Conger land Systems, Phoonix, Arizona, Tune 1987. 19, Jey M. Mendel. Use of higher order statist update. In Proceedings of SPIE Conference on A Processing, San Diego, California, August 1988. 26. Chaysostomos L. Nikias. Higher-order spectral analysis, In Simon Haykin, editor, vances in Speen Amaya and Array Processing, pages 326-365, Prentice Hal, Ine Englewood Cis New Jersey, 1992. 21. Cheysostomos I. Nikias and Mysore R, Raghaveet. Bispctrum estimation: « digital signal processing framework. Proceedings of the IEEE, 75(7:869-891, July 1987. 20. D.R. Biilinget. Am introduction to polyspectra, Annals of Mathemarical Siaristics, 36:1351 1374, 1965. 23. Chrysostomes L. Nikias. Higher Ine., Englewood Cliffs, New Jersey. 24. Murray Rosenblatt, Stationary Sequences and Random Fields sachusetts, 1985, rence on Mathematical Theary of Networks fics in signal processing and system theory: an \dvanced Algorithms and Architectwres for Signal ‘Order Moments in Digital Signal Processing, Prentice Hall, Biridituser, Cambridge, Mas- PROBLEMS 4A. (@) Determine the mean and autocorrelation function forthe random process fn] = vie) + 30 — 1) where nf] isa sequence of independent random variables with mean aud variance (Is z{n] stationary? 442 Random processes z{n] and yf ere defined by afr] = ln} + 3eln— vain + +3uln= 1 tent white noise processes each with variance equal t0 0.5. oP Ae these processes wide-sonse sta ate where »[n] and vafn) are independ {a) What are the autocorrelation functions of = and onary?220 Second Moment Analysis Chap. 4 422, What are the correlation functions forthe real and imaginary parts ofthe exponential correlation fonction in Table 4.27 What is the eross-correlation function? Determine also the complex spectral density function for the real and imaginary pans of the process and sketch its poles and zeros. 4.23, A certain real randam process is defined by Bin] = Acosuon + win} where A is a Gaussian random variable with mean zero and variance 0 and wf) is & white : noise process with variance a3, independent of A. (@) What is the corelation function of aia}? {b) Can the power spectrum of 2{n] be «fined? If so, what is the power spectral density function? (©) Repeat pars (a) and (b) in the ease when the cosine has an independent random pase uniformly distributed between — and 7, 424, (a) Derive a general expression for the corelation function of the random process defined by (4.103) and show that the process is wide-sense stationary if and only if the amplitudes satisfy the orthogonality condition (4.108) (b) By decomposing a real sinusoid with real random amplitude ané uniform phase into the sum oF two complex sinusoids, show that such a random process satisfies the orthogonality condition above and is therefore a stationary random process. (©) A sampled random square wave with discrete period P has the form AsgtinT ~ 7) | ain) where A is areal random variable, 7's the sampling interval, and > isa random delay pa rameter uniformly distributed over [0, PT} ad independent of A. This signal is comprised ff the fundamental frequency and only odd harmonies. Show that this periodic random process is also stationary. (@) Give an example of a random process in the form of (4.103) that does nor satisfy the orthogonality conditions (4,104), Recall that each component ofthe process is assumed 10 have uniform phase and amplitude distributed independent of phase 4.25. A sufficient condition forthe random process ln] = Ae = jal to be wide-sense stationary is that A and ¢ be independent and 6 be uniformly distributed “This condition also guarantees tha the essential requirement Efetrsztnal} = 0 is satisfied forthe complex random process, Show tha the foregoing condition is only suficient for the stationarity, but not necessary. In other words, show by counterexample that [Al and ( need not be independent, and tha if they are independent, the phase need not be uniformly distributed in order forthe random process to be stationary 4.26. By following a proceduse similar to that in Section 4.1.2, prove the positive semidefinite property for the correlation function of a continuous random process, Then by taking the function a0) in Table 4.7 to be an appropriate combination of continuous-time impulses, show that the | property (4.130) holds,270 Linear Transformations Chap. 5 (by What is the corelation function of the output? SA. A linear system Is defined by Tyn— Vj +2in) ~ ain — 1 ull = (a) Compute the fs four values of Rye) iit is known that Pf] = (b) What is Reylll for -3 <1 3? (©) What is the power spectral density function Sy( 155. A causl linear shift-invariant system is described by the difference equation oy 5 1 vis) Sula + gol 21 = aie TT nserve that he correlation and cross-conatin functions sts the difference equations Rll) = LAyall— 1+ gRoal 21> Rell —H Raft) ~ £Rglk— + gall) = Peel =H (Show tht I the input = 8 a white nie process wih it vine, then she slalom mere gatt-«((8)-G)) ‘where uff isthe unit step function. () The tontion above fs now used an input tthe second equation. Since the envi Poa recteite te sum of two exponentials, for I < 0 ts Teasonable (0 exsume that the response will be ofthe form nwea(t) te(f) 1 1 0 it is reasonable to assume that the only response will be Since there is n0 input f the transient response, which has @ rat=a(3) 4(G)2 1° ine second equation. WT agetem? Use this to find the z-transform similar form: ‘With these considerations, find the solution t {e) What is the system Function 24(2) of the origin of the correlation fonction R(t (a) What is the power spectrum Sy {e) Find the impulse response of the (5.13}45.15) to find the output com ©. 4:6, Find a genera closed-form expression forthe corelaton Function of the random Proves: tal eseribed by the first-order difference equation ain) + ayn t= sin) + bain = 1h ony? original system an relation funotion R(t). Ch 18 use the convolution relationships feck your answer with part when the input z{n] is white noise with variance os i |Problems a 5.7. A signal with correlation function RUS = (2) "4g applied to a linear shift-invariant system svth impulse response hn} = 6in} + 4fa ~ 1 {a> Compute the coreation faction of the outst {0) Compute the power specrum ofthe input. fo Compe the power spectrum ofthe owput 158. A random signal z{n] is passed through linea system with impulse response ‘in = Stn] ~ 25!0— 1) {@) Find the cros-corelation funetion between input and output Rey] i the input is white noise with variance of, (8) Find the correlation fonction ofthe output Rt {@) Find the output power spectral density Sy(e™) “The impulse response of a Linear shift-invariant system is given by ={cr ogg? 8s A ° otherwise [A white noise process with variance of = 1 i applied to this system. Call the inpu to the System sn} and the evipat ln {ap Whats the crss-conslaton Fonction Ray? Sketch this function. {8) What is the corelation function Ry{l} ofthe output? Sketch this neatly. |5:10. A real random process with correlation Function Rall = 208)" invariant system whose difference equation is is epplied to a inear shi yin) = 0.5yln — 1+ afr) a) What isthe complex spectral density function of the output Sy(2)? (b) ‘What is the corelation function of the output Fy[U7? S11. A random process z[1} consists of independent random variables each with uniform deasity ; fo -texstt s0n={ 5 aaevae applied to linear shif-ivarant system sith impulse response _fq@r ze wan {P02 Let the output process be denoted by pln (a) Compute Fy (b) What is Ri? (€) What is 3,(2)? Use this 10 compute $,(e 5.12 A linear shif-invarian system has the impulse response ‘n} = 28n} + 8fn ~ 1) Ate 21 “This process i 0,é 276 Linear Transformations Chap. 5 ‘Thus if () is defined as fu) = In|") then it is equivalent so show that Freed] dir < 00 (@) Define the function atwa{ tm 20 ° otherwise | Show that : | eda >? fated | Further show that since [suerte <2 / i, this implies that . | [ores [omc [Assume that Hg(2) has no 2eros on the unit circle) (@) Finally, show from parts (a) and (b) that 2 [inode [painter <0 1527, Factor the following complex spectral density functions into minimam- and maximum phase on L -16 SOm B+ Tet oo delet +34 SO Toe Bet 15.28. A random process has the complex spectral density anton are the poles and (a) Factor this fonction into minimum- and maximum-phase terms. Wt