上海交通大学自适应信号处理课程资料.7z

clc; close all clear variables; length=100; % sequence length iteration_num=100; % iteration number a1=0.1; % AR coefficient a2=-0.8; var_v=0.27;% noise variance sd_v=sqrt(var_v); mu=0.05; u=zeros(length+3,1); f=zeros(length+3,1); epsilon1=zeros(length+3,1); epsilon2=zeros(length+3,1); g=zeros(length+3,1); J=zeros(length+3,1); weights=zeros(2,length+3); gsum = zeros(length+3,1); for k=1:iteration_num % number of iteration W=zeros(2,length+3); for n=3:length+3 u(n)=a1*u(n-1)+a2*u(n-2)+randn(1)*sd_v; f(n)=u(n)-W(1,n-1)*u(n-1)-W(2,n-1)*u(n-2); epsilon1(n)=a1-W(1,n-1); epsilon2(n)=a2-W(2,n-1); W(:,n)=W(:,n-1)+mu*f(n)*[u(n-1);u(n-2)]; end g=g+f.^2; % accumulate squared error of estimation end g=g/iteration_num; % accumulate squared error of estimation for n=3:length+3 % theoretical cost function J(n)=(1-var_v*(1+mu/2))*(1-mu)^(2*n)+var_v*(1+mu/2); end index=1:length+3; subplot(2,2,1) plot(index,(abs((fft(f)))).^2) xlim([0 length+3]) title('Power Spectral of f') xlabel('Frequency') ylabel('Power Level') subplot(2,2,2) plot(index,(abs((fft(epsilon1)))).^2) xlim([0 length+3]) title('Power Spectral of \epsilon_1') xlabel('Frequency') ylabel('Power Level') subplot(2,2,3) plot(index,(abs((fft(epsilon2)))).^2) xlim([0 length+3]) title('Power Spectral of \epsilon_2') xlabel('Frequency') ylabel('Power Level') figure plot(index,g(1:length+3),index,J) xlim([0 length+3]) ylim([0 1]) legend('Experimental Results','Theoretical Results') xlabel('Sequence length') ylabel('Squared Error') title({'Comparison Plot of Theoretical Learning Curve';... 'to Observed MonteCarlo Results'})