SPEECH SIGNAL ENHANCEMENT IN HEAVY NOISE INDUSTRIES

Abstract

Audio signal in speech communication is corrupted by an additive random noise where there is a noisy environment. These noisy surroundings might be a moving car, train, factory, or a noisy telephone channel. Since noise is random and varying continuously, we need to estimate the noise at every instant to remove it from the desired signal. There are many schemes for noise reduction but the most efficient scheme to accomplish noise cancellation is to employ adaptive filters. In speech signal enhancement there are two categories of algorithms in which either a single microphone or multiple microphones are employed to clean up the noisy signal. In this paper, Matlab simulations of different adaptive algorithms and comparison of their performances for noise cancellation in a noisy environment, specifically industrial noise, is carried out. A comparative analysis of different adaptive filters in the presence of a single and dual microphone is provided. A robust voice activity detector (VAD) is incorporated in the single channel speech enhancement.