Two parallel filters with one-dimensional non-local mean method for noise cancellation in audio signals

Abstract EN

Signal denoising is an important process and should be implemented before any advanced process on the audio signals.

This paper proposes a two -stage detection method to detect the noisy samples effectively.

In the first stage, two parallel filters are primarily implemented to determine the common noisy samples.

In the second stage, intensity difference based method is employed in order to determine the truly noisy samples by filtering the common noisy samples.

At the end, one-dimensional non local mean filter is deployed for estimating the truly noisy samples rather than estimating all the samples in the signal.

Extensive simulation results show that the proposed method delivers superior results at high and low noise rates with respect to the other well-known denoising methods.

Moreover, the suggested method is fast and consistent against the values of the suggested parameters.