Objective Evaluation Measures for Speech Enhancement for Non Stationary Noise Environments

Abstract

A new speech enhancement algorithm was once proposed in this paper with an intention of decreasing the non-stationary noises introduced on the easy speech signals. Suppression of non-stationary noise in a imperative issue. Because, the traits of noises are one of a kind shape noise to noise as nicely as varies with admire to the environment. To solve this issue, a novel non-stationary noise suppression mechanism is proposed in this paper based on the Sub band Adaptive Filtering. The main advantage with the accomplishment of SAF is faster convergence and also better quality achievement. Along with SAF, a noise classification mechanism additionally proposed in this paper to limit the extra computational complexity in the noise identification. Extensive simulations are carried out over the proposed mechanism thru distinctive speech indicators with different noise at exclusive SNRs. The performance evaluation is carried out through the performance metrics, PESQ and WSS and reveals the outstanding performance of proposed mechanism.