linear convolution

Table of Contents

A linear convolution is the sum of the product of two sequences after one is reversed and progressively shifted.

(f[n]g[n])=(k=)()f[x]g[nk]

Here, we are basically assuming that each signal is infinitely padded with zeros.

This operation commutes.

For example, (123)(45)=((34)(24+35)(14+25)(15))

The most important use of the linear convolution is to compute the output of a linear time invariant system.

It is possible to compute the linear convolution fast using the FFT. We need to pad each sequence to (at least) length N+M1 with zeros at the end. This basically removes the overlap in the circular convolution.

This number is relevant because its the output length of a lin. conv. between signals of length n, m.

1. Further reading:

Back to index