exadapsp.m

these code are about adaptive filter.

%          --------------------------------
%  +++++    NOTE: THIS IS A MATLAB PROGRAM    +++++++
%          --------------------------------   


% ****  This Program solves the Problem in Fig 8.7 (p 169) in notes   ****
%                   (Same as Fig 9.19 of Haykin)  

% (i.e. can be used to produce Fig 8.11 (Haykin Figure 9.23) )
% 

w = 3.1 ; ite = 1500 ; mu = 0.025 ; ens = 200 ;

%
% h   ==> channel response
% x   ==> channel input
% del ==> used to delay input and
%           obtain the desired response
% u   ==> tap weight vector
% cw  ==> FIR filter weights
%

h = 0.5*(1.0 + cos(2*pi*([0:4]' - 2)/w)) ;
h(1) = 0 ; h(5) = 0 ;

enmse = zeros(1, ite);

for  kk = 1:ens
  del = zeros(8,1) ;
  x = zeros(5,1) ;
  cw = zeros(11,1) ;
  u = zeros(11,1) ;

for k = 1:ite
  del(2:8) = del(1:7);
  x(2:5) = x(1:4);
  
  % calculate the Bernoulli input
  x(1) = 2*floor(rand(1,1)+0.5) - 1 ;
  del(1) = x(1) ;
  
  % calculate the tap inputs
  y = h'*x + randn(1,1)*sqrt(0.001) ;
  u(2:11) = u(1:10) ; u(1) = y ;
  
  err = del(8) - cw'*u ;
  cw = cw + mu*err*u;
  
  mse(k) = err*err ;
  end ;  % end of k loop %

  % now calculate the ensemble average
  enmse = enmse + mse/ens ;
end ;   % end of kk loop %

%
% Now plot one (last) realization and also
% ensemble average (learning curve).
%

plot([1:ite], log10(mse), [1:ite], log10(enmse)) ;
grid ;