Speech enhancement by speech intelligibility index in sensor network

Abstract

Speech enhancement aims to improve speech quality by using various algorithms. The main objective of enhancement is to improvement in intelligibility and overall perceptual quality of degraded speech signal using audio signal processing techniques. In the field of speech enhancement, enhancing of speech means degraded by noise in its wide range of applications such as mobile phones, VoIP, teleconferencing systems etc. In general there are three different methods used to estimate speech intelligibility. Namely, Speech Intelligibility Index (SII), Speech Transmission Index (STI) and Articulation Index (AI). Here it is proposed that SII is most robust physical measure and the comparison between Speech Intelligibility index in presence of stationary noise (White Gaussian Noise) and non-stationary noise (Speech noise) is done. Simulation result shows that SII is better in presence of non-stationary noise (a female voice of sampling frequency 16 KHz). Here two wideband speech signals are considered for performance evaluation since it brings the improvement over traditional narrowband such as increases the intelligibility and enables the spatial auditory displays etc. The speech signal are generated and simulated with the MATLAB environment. The real time speech signals are recorded with the help of acoustic sensor present inside the microphone.