masconv

The most easily understood program of convolution

Description

This program to perform convolution has been written keeping the ordinary polynomial multiplication in mind and in my opinion is the most easily understood program among the ones I have seen on the web or written by other fellow workers although it may not be the best program in its design. The user may get help upon usage by typing help masconv at the command prompt.

When the user gives the program two vectors (sequences) to be convolved, the program first initialize a temporary matrix of size having number of rows according to the length of the first vector and number of columns one less than the length of both the vectors (as it is required for the convolution sum to have this much length) using the zeroscammand. The program then make use of two for loops to replace the zero values in the temporary matrix at desirable places by simple multiplication of the members of the two vectors. Once the temporary matrix is completed, the program then use the sum command of Matlab to find the final convolution sum and displays the result.

The user can compare the results of this program with the built-in function conv. Any suggestions or criticisms are welcome.

• MATLAB release: MATLAB 6.5 (R13)

[Imran Zaka]

 

[Rashid Saleem]

Excellent!

[Muhammad Saeed]

 

[John D'Errico]

This code is purely for expository purposes, as conv will be preferred always. As a conv tutorial, this does show nicely the basic idea of how one might write a convolution, except one should not (in practice) generate the entire table first, and THEN do the sum. This will be an inefficient and unnecessary use of memory. I can forgive inefficiency for expository purposes, as long as the reader is made aware of the difference. As such, a few comments to that effect would possibly increase my rating.

It might also be of interest to show the reader a more
efficient version. In some quick tests, with length(a) = 1000 and length(b) = 100, the version below is 8x faster, but with comments it should be as understandable as the doubly looped version.

na = length(a);
nb = length(b);
b=b(:);
result = zeros(na+nb-1,1);
for m = 1:na
  result(m:(m+nb-1)) = result(m:(m+nb-1)) + a(m)*b;
end

[Ashar Ahmad]

 

[salman rashid]

 This code is purely for expository purposes, as conv will be preferred always. As a conv tutorial, this does show nicely the basic idea of how one might write a convolution, except one should not (in practice) generate the entire table first, and THEN do the sum. This will be an inefficient and unnecessary use of memory. I can forgive inefficiency for expository purposes, as long as the reader is made aware of the difference. As such, a few comments to that effect would possibly increase my rating.

[sreenivas Puthiya Veettil]