Overlap-Save Method

The overlap-save method is a signal processing technique for efficiently computing the convolution of a very long signal with a finite impulse response filter. Instead of filtering the whole signal at once, it splits the input into overlapping segments, converts the linear convolution of each segment into a circular convolution, and evaluates that circular convolution using the fast Fourier transform, discarding the parts of each segment's output that were corrupted by the overlap. Because FFT-based circular convolution is cheaper than direct convolution for long filters, the method needs far fewer multiplications per output sample than direct convolution would, which makes it useful for real time filtering of long data streams.