A Review on Beamforming Approaches for Speech Enhancement

Abstract

The main aim of this paper is to discuss speech enhancement techniques using delay & sum, complex Gaussian Mixture Model and Minimum variance distortion less response (MVDR) beam former. Using spatial filtering in microphone arrays results in increasing the required signal in the existence of environmental noise sources. These techniques i.e Delay and Sum Beam former(DSB), Minimum Variance Distortion less Response (MVDR) beam former and Complex Gaussian Mixture Model (CGMM) are implemented using multiple channel enhancement and the performance of these techniques were analyzed for different noises under various SNR  levels. Improved speech qualities are optimized in time domain and frequency domain using beamformer approaches. The results shows that, rather than using delay and sum beam former technique the speech quality is improved with Minimum Variance Distortion less Response Beam former. The performance of different methods are analyzed using objective performance measures Signal to Noise Ratio (SNR), Perceptual Evaluation of Speech Enhancement (PESQ).