vrls.m

非线性自适应滤波器MATLAB程序

%VRLS Volterra RLS algorithm
%
%   'ifile.mat' - input file containing:
%      Nr - members of ensemble
%      dim - iterations
%      Sx - standard deviation of input
%      Sn - standard deviation of measurement noise
%      lambda - exponential weighting factor
%      
%   'ofile.mat' - output file containing:
%      MSE - mean-square error


clear all		% clear memory
load ifile;		% read input variables

for j=1:Nr
   n=Sn*randn(dim,1);        % noise at system output 
   x=Sx*randn(dim,1);       % input signal 
   xl1=zeros(dim,1); xl2=xl1;
   xl1(2:dim)=x(1:dim-1); xl2(3:dim)=x(1:dim-2);
   d=zeros(dim,1);
   d=-.76*x-xl1+xl2+.5*x.^2+2*x.*xl2-1.6*xl1.^2+1.2*xl2.^2+.8*xl1.*xl2+n; ...
     % unknown system output
   w=zeros(9,dim);           % initial coefficient vector
   uxl=[x xl1 xl2 x.^2 x.*xl1 x.*xl2 xl1.^2 xl1.*xl2 xl2.^2]'; % input vectors
   
   Sd=eye(9);
   for i=1:dim
      elinha(i)=d(i)-w(:,i)'*uxl(:,i)+n(i);  % error sample
      psi=Sd*uxl(:,i);
      Sd=(1/lambda)*(Sd-(psi*psi')/(lambda+psi'*uxl(:,i)));
      w(:,i+1)=w(:,i)+elinha(i)*Sd*uxl(:,i); % new coefficient vector
      y(i)=w(:,i+1)'*uxl(:,i);               % output sample
      e(i)=d(i)-y(i)+n(i);
   end
   mse(j,:)=e.^2;
end

MSE=mean(mse);

save ofile MSE;