广西大学《数字信号处理》课件-Chapter5 Finite-Length Discrete Transform.ppt

DSP group chap5 * Sect. 5.3.3 DTFT from DFT by Interpolation Let DSP group chap5 * Sect. 5.3.4 Sampling the DTFT DSP group chap5 * (a) M=N Sect.5.3.4 Sampling the DTFT (b) MN: (c) MN: x[n] is preserved to be unchanged. For a finite-length sequence x[n],0?n?M?1 For a infinite-length sequence x[n],0?n?M?1 DSP group chap5 * DTFT Real line Sampling No time aliasing IDFS MN 0? n ? 9 0? k ? 15 Sect. 5.3.4 Sampling the DTFT--- Length of sequence M is less than sampling point N y[n] x[n] DSP group chap5 * DTFT Real line Sampling time aliasing IDFS M ?N 0? n ? 9 0? k ? 5 Sect.5.3.4 Sampling the DTFT--- Length of sequence M is greater than sampling point N y[n] x[n] Problem: Is there any wrong in the above figures? DSP group chap5 * Problem1: if sequence x[n] is an infinite-length sequence, how can we get the numerical computation? Problem2: Since the DFT can be thought to be the sampling of DTFT, how can we obtain the DTFT X(e j?) from its samples X[k]? Sect.5.3.4 Sampling the DTFT The second problem See Section 5.3.3; The first problem help us to compute the numerical value of X(e j?) correctly. DSP group chap5 * Ex. 5.6 Find y[n] which is the IDFT of Y[k]. -------x[n] can be recovered from y[n] Solution: Sect.5.3.4 Sampling the DTFT Solution: -------x[n] cannot be recovered from y[n] Ex. 5.6-2 DSP group chap5 * Sect.5.3.4 Sampling the DTFT DSP group chap5 * Sect.5. 4 Circular convolution Read material in sect.2.3 “circular operation” Linear convolution Assume two length-N sequences, g[n],0?n?N?1, and h[n],0?n?N?1 Circular convolution DSP group chap5 * yC[n] = g[n] h[n] (5.52) N Denoted as Sect.5. 4 Circular convolution Computation: DSP group chap5 * ----The circular convolution is commutative, i.e. g[n] h[n] = h[n] g[n] (5.53a) N N Sect.5. 4 Circular convolution (g[n] h[n]) w[n] = g[n] (h[n] w[n]) (5.53b) N N N N Example: Determine the 4-point circular con