DSP_MATLAB.zip

% exa160301b, 研究LMS算法的性能，求学习曲线。应用模型是系统辨识，见exa160501； clear all; N1=1000; N=100; M=15; b = fir1(M-1,0.5); % FIR system to be identified mu1 = 0.001; % LMS step size mu2 = 0.01; % LMS step size for k=1:N x = randn(1,N1); % Input to the filter noise = 0.1*randn(1,N1); % Observation noise signal d = filter(b,1,x)+noise; % Desired signal ha1 = adaptfilt.lms(M,mu1); [y1,e1] = filter(ha1,x,d); e21(k,:)=e1; ha2 = adaptfilt.lms(M,mu2); [y2,e2] = filter(ha2,x,d); e22(k,:)=e2; end E_mu1 = sum(e21.^2,1)./N; E_mu2 = sum(e22.^2,1)./N; figure(1) plot(E_mu1);grid; hold on; plot(E_mu2); hold on; xlabel('Number of interations') ylabel('Mean square error for mu1,mu2') N=10*N; for k=1:N x = randn(1,N1); % Input to the filter noise = 0.1*randn(1,N1); % Observation noise signal d = filter(b,1,x)+noise; % Desired signal ha1 = adaptfilt.lms(M,mu1); [y1,e1] = filter(ha1,x,d); e21(k,:)=e1; ha2 = adaptfilt.lms(M,mu2); [y2,e2] = filter(ha2,x,d); e22(k,:)=e2; end E_mu1 = sum(e21.^2,1)./N; E_mu2 = sum(e22.^2,1)./N; plot(E_mu1,'r');grid; hold on; plot(E_mu2,'r'); rd=xcorr(d,'biased'); rx=xcorr(x,'biased'); rxd=xcorr(d,x,'biased'); RX=toeplitz(rx(N1:(N1+(M-1)))); e_min=rd(N1)-rxd(N1:(N1+(M-1)))*inv(RX)*rxd(N1:(N1+(M-1)))'; e_min