Enhancement of Noisy Speech with Low Speech Distortion Based on Probabilistic Geometric Spectral Subtraction

TL;DR

This paper introduces a probabilistic geometric spectral subtraction method for noisy speech enhancement that effectively reduces noise while preserving speech quality by using a confidence parameter in the gain function.

Contribution

It proposes a novel spectral subtraction approach with a confidence parameter to improve noise suppression and speech quality preservation.

Findings

Effective noise reduction demonstrated in simulations

Prevents speech distortion during enhancement

Outperforms traditional spectral subtraction methods

Abstract

A speech enhancement method based on probabilistic geometric approach to spectral subtraction (PGA) performed on short time magnitude spectrum is presented in this paper. A confidence parameter of noise estimation is introduced in the gain function of the proposed method to prevent subtraction of the overestimated and underestimated noise, which not only removes the noise efficiently but also prevents the speech distortion. The noise compensated magnitude spectrum is then recombined with the unchanged phase spectrum to produce a modified complex spectrum prior to synthesize an enhanced frame. Extensive simulations are carried out using the speech files available in the NOIZEUS database in order to evaluate the performance of the proposed method.