Signal processing part I_cn_85课件.ppt

Signal processing:信号处理:Part I : Stationary signal稳态信号;;;;;;;;;;;;;;;;;Convolution between 2 signals f and g 两个信号f 和g的卷积 Is a mathematical operator which calculate the amount of overlap between f and a reverse or translated version of g. A convolution is a kind of moving average. 这是一个数学运算,表示 f 和 经过反褶或者平移的g的重叠量的积分, 卷积就是一种移动平均. Convolution in time domain will become multiplication in frequency domain. 时域的卷积变成频域的乘法运算;;;;;;;;;;;;;;;3.2 Sampling frequency fs -- 采样频率 fs is also the inverse of the sampling frequency fs. We have seen the consequence of sampling a continuous signal. 是采样频率fs.的倒数, 我们已经看到采样连续信号序列 If the sampling frequency is too low ( too large) the frequency spectrum overlaps, and become corrupted. 如果采样频率太低(太大), 频谱就会发生重叠错误 ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? ;;;;3.5 Final DFT 最终的 DFT Recall the following writing of the DFT: 前面讲过的DFT: As soon as the frequency spectrum is also sampled : f = m 频谱经过采样: f = m Or 或者 Therefore : 因此:;3.6 FFT Fast Fourier Transform 快速傅立叶变换FFT The FFT is just a quick way of computing the DFT by having N of being of the form N=2p. FFT是一种快速DFT算法, N 为 N=2p The number of sample values N is 64, 128, 256,1024, 2048….. 采样数N 为: 64, 128, 256,1024, 2048….. With the DFT, N2 complex multiplications must be calculated to get from N frequency components N samples. Whereas with FFT only N*log2N multiplications are enough. Which is much faster. DFT的计算量为N2，是频率个数N和采样数N的乘积, 如果是FFT的话, 计算量只有N*log2N FFT was invented in 1965 by James W.Cooley and John W. Tukey. FFT 由 James W.Cooley 和 John W. Tukey 在1965年发明. ;IV. Quantization 量化 Up to now the signal has been sampled along the time or frequency axis. It also need to be digitized in terms of amplitude which is called the quantization. 到目前信号已经沿着时间轴和频率轴被采样, 幅值的数字化就叫做量化; In the binary system the number of steps for the quantization is equal to 2n where n