[go: up one dir, main page]
Scribd
Upload
70 views1 page

Dft-Lms Filter

This document describes a DFT-LMS filter algorithm that takes in a signal with noise and estimates the noise to filter it out. It does this by: 1) Taking the FFT of blocks of the noise signal to estimate it in the frequency domain. 2) Subtracting this estimated noise from the original signal to get an error signal. 3) Updating the filter weights using the error signal and power estimates to minimize the estimated noise for the next block. 4) Outputting the filtered signal with the estimated noise removed.

Uploaded by

api-381840243
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
70 views1 page

Dft-Lms Filter

This document describes a DFT-LMS filter algorithm that takes in a signal with noise and estimates the noise to filter it out. It does this by: 1) Taking the FFT of blocks of the noise signal to estimate it in the frequency domain. 2) Subtracting this estimated noise from the original signal to get an error signal. 3) Updating the filter weights using the error signal and power estimates to minimize the estimated noise for the next block. 4) Outputting the filtered signal with the estimated noise removed.

Uploaded by

api-381840243
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

DFTLMSFilter

Input:miu=stepsizeN=lengthofweighty_t=signal+noise(originallength)x_t=noise(originallength)r=forgettingratioP=
initialPower

function output = dftLmsFilt(miu,N,y_t,x_t,r,P)


W = zeros(2*N,1);
output = y_t;

a = length(x_t);
Blocks = a/N;

for k =1:Blocks - 1
%taking first N noise and fft
X = fft([x_t((k-1)*N + 1: (k+1)*N)],2*N);
x_estimate = ifft(X.*W);
%since FFT is symmetric only take last half
x_estimate = x_estimate(N+1:2*N,1);

%take first from N to N+K(lost first N)


y_N = y_t(k*N + 1: (k+1)*N);

%calculate error signal


output(k*N + 1: (k+1)*N,1)=y_N - x_estimate;

%calculate weights!

E_error = fft([zeros(N,1);output(k*N+1:(k+1)*N)],2*N);
%estimate power
P = r*P+(1-r)*abs(X).^2;
Pinv = 1./P;

%estimate gradient(min of estimated noise)


E = ifft(Pinv.*conj(X).*E_error,2*N);
%take only first half
E = E(1:N);

%update the weights [E;zeroes(N,1)] because we take fft to put it


%to
W = W + miu*fft([E;zeros(N,1)],2*N);

end
% make sure it is real
output=real(output(:));
end

Not enough input arguments.

Error in dftLmsFilt (line 11)


W = zeros(2*N,1);

PublishedwithMATLABR2017b

You might also like