Scribd
Upload
122 views

Overlap Save Convolution

The document describes the overlap-add and overlap-save methods for efficiently computing the convolution of a long signal with a finite impulse response (FIR) filter. The overlap-add method divides the long signal into blocks, computes the convolution of each block with the filter, and then overlaps and adds the results to obtain the final convolution. The overlap-save method uses blocks that overlap, computing the convolution in the frequency domain after zero-padding and discarding aliased portions of the result. Pseudocode is provided for both methods.

Uploaded by

lachuns123
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
122 views

Overlap Save Convolution

The document describes the overlap-add and overlap-save methods for efficiently computing the convolution of a long signal with a finite impulse response (FIR) filter. The overlap-add method divides the long signal into blocks, computes the convolution of each block with the filter, and then overlaps and adds the results to obtain the final convolution. The overlap-save method uses blocks that overlap, computing the convolution in the frequency domain after zero-padding and discarding aliased portions of the result. Pseudocode is provided for both methods.

Uploaded by

lachuns123
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 4

Theory:

%
% Overlap Add Method:
% The overlapadd method is an efficient way to evaluate the discrete convolution of a very long signal with a finite
impulse response
% (FIR) filter where h[m] = 0 for m outside the region [1, M].The concept here is to divide the problem into
multiple convolutions of h[n]
% with short segments of x[n], where L is an arbitrary segment length. Because of this y[n] can be written as a sum
of short convolutions.
%
% Algorithm:
%
% The signal is first partitioned into non-overlapping sequences, then the discrete Fourier transforms of the
sequences are evaluated by
% multiplying the FFT xk[n] of with the FFT of h[n]. After recovering of yk[n] by inverse FFT, the resulting output
signal is reconstructed by
% overlapping and adding the yk[n]. The overlap arises from the fact that a linear convolution is always longer than
the original sequences. In
% the early days of development of the fast Fourier transform, L was often chosen to be a power of 2 for efficiency,
but further development has
% revealed efficient transforms for larger prime factorizations of L, reducing computational sensitivity to this
parameter.
% A pseudo-code of the algorithm is the following:
%
% Algorithm 1 (OA for linear convolution)
% Evaluate the best value of N and L
% H = FFT(h,N)
(zero-padded FFT)
% i=1
% while i <= Nx
%
il = min(i+L-1,Nx)
%
yt = IFFT( FFT(x(i:il),N) * H, N)
%
k = min(i+N-1,Nx)
%
y(i:k) = y(i:k) + yt (add the overlapped output blocks)
%
i = i+L
% end
%
% Circular convolution with the overlapadd method
%
% When sequence x[n] is periodic, and Nx is the period, then y[n] is also periodic, with the same period. To
compute one period of y[n],
% Algorithm 1 can first be used to convolve h[n] with just one period of x[n]. In the region M ? n ? Nx, the
resultant y[n] sequence is correct.
% And if the next M ? 1 values are added to the first M ? 1 values, then the region 1 ? n ? Nx will represent the
desired convolution.
% The modified pseudo-code is:
%
% Algorithm 2 (OA for circular convolution)
% Evaluate Algorithm 1
% y(1:M-1) = y(1:M-1) + y(Nx+1:Nx+M-1)
% y = y(1:Nx)
% end
%
clc;
clear all;
Xn=input('Enter 1st Sequence X(n)= ');

Hn=input('Enter 2nd Sequence H(n)= ');


L=input('Enter length of each block L = ');
% Code to plot X(n)
subplot (2,2,1);
stem(Xn);
xlabel ('n---->');
ylabel ('Amplitude ---->');
title(' X(n)');
%Code to plot H(n)
subplot (2,2,2);
stem(Hn,'red');
xlabel ('n---->');
ylabel ('Amplitude ---->');
title(' H(n)');
% Code to perform Convolution using Overlap Add Method
NXn=length(Xn);
M=length(Hn);
M1=M-1;
R=rem(NXn,L);
N=L+M1;
Xn=[Xn zeros(1,L-R)];
Hn=[Hn zeros(1,N-M)];
K=floor(NXn/L);
y=zeros(K+1,N);
z=zeros(1,M1);
for k=0:K
Xnp=Xn(L*k+1:L*k+L);
Xnk=[Xnp z];
y(k+1,:)=mycirconv(Xnk,Hn); %Call the mycirconv function.
end
p=L+M1;
for i=1:K
y(i+1,1:M-1)=y(i,p-M1+1:p)+y(i+1,1:M-1);
end
z1=y(:,1:L)';
y=(z1(:))'
%Code to plot the Convolved Signal
subplot (2,2,3:4);
stem(y,'black');
xlabel ('n---->');
ylabel ('Amplitude ---->');
title('Convolved Signal');
% Add title to the Overall Plot
ha = axes ('Position',[0 0 1 1],'Xlim',[0 1],'Ylim',[0 1],'Box','off','Visible','off','Units','normalized', 'clipping' , 'off');
text (0.5, 1,'\bf Convolution using Overlap Add Method ','HorizontalAlignment','center','VerticalAlignment', 'top')
Theory
%
% Overlap Save Method
%

% In this method, the size of the input data blocks is N=L+M-1 and the DFTs and the IDFTs are of length L. Each
Data Block consists of the
% last M-1 data points of the previous block followed by L new data points to form a data sequence of length
N=L+M-1.An N point DFT is computed
% for each data block. The impulse response of the FIR filter is increased in length by appending L-1 zeros and an
N-point DFT of the sequence is
% computed once and stored. The multiplication of the N-point DFTs for the mth block of data yields
%
Ym(k)=h(k)Xm(k).
% Since the data record is of length N, the first M-1 points of Ym(n)are corrupted by aliasing and must be discarded.
The last L points of Ym(n)
% are exactly the same as the result from linear convolution. To avoid loss of data due to aliasing, the last M-1
points of each data record are
% saved and these points become the first M-1 data points of the subsequent record. To begin the processing, the
first M-1 point of the first record
% is set to zero. The resulting data sequence from the IDFT are given where the first M-1 points are discarded due to
aliasing and the remaining L
% points constitute the desired result from the linear convolution. This segmentation of the input data and the fitting
of the output data blocks together
% form the output sequence.
%
%
clc;
clear all;
x=input('Enter 1st sequence X(n)= ');
h=input('Enter 2nd sequence H(n)= ');
L=input('Enter length of each block L = ');
% Code to plot X(n)
subplot (2,2,1);
stem(x);
stem(x,'blue');
xlabel ('n---->');
ylabel ('Amplitude ---->');
title('X(n)');
%Code to plot H(n)
subplot (2,2,2);
stem(h);
stem(h,'black');
xlabel ('n---->');
ylabel ('Amplitude ---->');
title(' H(n)');
% Code to perform Convolution using Overlap Save Method
M=length(h);
lx=length(x);
r=rem(lx,L);
x1=[x zeros(1,L-r)];
nr=(length(x1))/L;
h1=[h zeros(1,L-1)];
for k=1:nr
Ma(k,:)=x1(((k-1)*L+1):k*L)
if k==1
Ma1(k,:)=[zeros(1,M-1) Ma(k,:)];
else
Ma1(k,:)=[Ma(k-1,(L-M+2):L) Ma(k,:)];

end
Ma2(k,:)=ifft(fft(Ma1(k,:)).*fft(h1));
end
Ma3=Ma2(:,M:(L+M-1));
y1=Ma3';
y=y1(:)'
% Representation of the Convoled Signal
subplot (2,2,3:4);
stem(y,'red');
xlabel ('n---->');
ylabel ('Amplitude ---->');
title ('Convolved Signal');
% Add title to the Overall Plot
ha = axes ('Position',[0 0 1 1],'Xlim',[0 1],'Ylim',[0 1],'Box','off','Visible','off','Units','normalized', 'clipping' , 'off');
text (0.5, 1,'\bf Convolution using Overlap Save Method ','HorizontalAlignment','center','VerticalAlignment', 'top')

You might also like