analogy.m

对示波器采集数据进行AR谱估计

%Conputer experiments on adaptive equalization  Experiment 2: LMS algorithm
clear;close all;clc;
load('TEK00007.DAT');%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%load('TEK00009.DAT');%noise
%signal=TEK00011';%load('TEK00004.DAT') 10004*1
signal=TEK00007';
noise=zeros(1,4000);
W1=zeros(1,2200);
W2=zeros(1,2200);
W3=zeros(1,2200);
W4=zeros(1,2200);
W5=zeros(1,2200);
W6=zeros(1,2200);
for i=1:4000
    noise(1,i)=signal(1,6000+i);
end
%noise=TEK00009';%load('TEK00003.DAT')  10004*1
N=3990;
%delta=7;
u3=[2.35 0.025 0.0075];
wk_mean=zeros(4,6);%****************************
err_mean=zeros(4,7993);
for ii=1:3
 %  w=3.3;
    u=u3(ii);             % w the factor of amplitude distortion;u step length
    wk=zeros(100,6);      % wk=zeros(N,11);%Weight vector   ZEROS([M,N]) is an M-by-N matrix of zeros***********************************
    err=zeros(100,N-7);
  %  power_v=0.001;   
  %  hn=[0,0.5*(1+cos(2*pi/w*((1:3)-2)))];%The impulse response of the channel
  %  Am=sqrt(power_v);      % Power of additive gaussian white noise 
    z=zeros(1,11000);
    y=zeros(1,11000);  
   for nn=1:100
  %     bernuli=randsrc(1,N);% Generate Bernoulli sequence   RANDSRC(M,N) generates an M-by-N random bipolar matrix                      1*1000
  %     vn=Am*randn(1,N+3);%Additive gaussian white noise   RANDN(M,N) and RANDN([M,N]) are M-by-N matrices with random entries          1*1003
  %     signal_in=[zeros(1,10),conv(bernuli,hn)+vn];%The input signal of adaptive transversal equalization   C = CONV(A, B) convolves vectors A and B.  The resulting.   1*1013
  %     bernuli_delta=[zeros(1,delta),bernuli];  %    1*1007
        for m=1800:N-7
            alc_data=noise(m+5:-1:m);  %signal+noise 
%            alc_data=noise(m); %level+noise 
            z(m)=alc_data*wk(nn,:)';        %trasmation of noise
            err(nn,m)=signal(m)-alc_data*wk(nn,:)';  %
            wk(nn,:)=wk(nn,:)+u*err(nn,m)*noise(m+5:-1:m); %
%            wk(nn,:)=wk(nn,:)+u*err(nn,m)*noise(m); %
            W1(1,m)=wk(nn,1);
            W2(1,m)=wk(nn,2);
            W3(1,m)=wk(nn,3);
            W4(1,m)=wk(nn,4);
            W5(1,m)=wk(nn,5);
            W6(1,m)=wk(nn,6);
        end
    end
    wk_mean(ii,:)=sum(wk)/100;%
    err_mean(ii,:)=sum(err.*err)/100;%
end
subplot(3,1,1);
stem(wk_mean(1,:));title('u=0.0075');axis([0 11 -1 1.8]);grid on;
subplot(3,1,2);
stem(wk_mean(2,:));title('u=0.025');axis([0 11 -1 1.8]);grid on;
subplot(3,1,3);
stem(wk_mean(3,:));title('u=0.075');axis([0 11 -1 1.8]);grid on;
figure;
semilogy(1:993,err_mean(1,1:993),1:993,err_mean(2,1:993),1:993,err_mean(3,1:993));
xlabel('Number of iterations n');ylabel('Ensemble averange square error');
legend('u=0.0075','u=0.025','u=0.075');title('LMS algorithm');
%
%%%%%%%后面预处理
%sk=zeros(1,7993);
%>> for j=1:7993
%   sk(1,j)=err(1,j);
%   end