A new FMH filter algorithm for removing impulse noise

Abstract

The proposed novel algorithm operates on the output of the FMH filter and reduces the averaged impulse type noise in speech/audio signals effectively and reduces jitter. The algorithm estimates the likelihood of the sample under inspection is corrupt relative to the neighboring samples and replaces a sample detected as corrupted by a value based on the neighboring samples. In the proposed algorithm there is no need to preset a threshold value to detect the averaged-impulse-noise. The algorithm has the advantage of being configurable to the type of noise impulses in the sample, as the threshold used to detect noise impulse varies to suit the signal.