Performance Evaluation of STSA Based Speech Enhancement Techniques for Speech Communication System

Abstract

Researchers present noise suppression model for reducing the spectral effects of acoustically added noise in speech. Background noise which is acoustically added to speech may decrease the performance of digital voice processors that are used for applications such as speech compression, recognition, and authentication. [6, 7] In proposed paper different types of Short Time Spectral Amplitude (STSA) [1, 17] based methods are explained to decrease the noise. Spectral subtraction gives a computationally efficient, processor- independent approach to effective digital speech analysis. But as a result of artifact, another synthetic noise may be produced by algorithm that is called musical noise. In spectral subtraction methods, there is shown less trade-off between residual and musical noise so the quality and intelligibility of signal is not maximized at required level. [8] To overcome from the problem of musical noise, wiener filter and statistical based model methods are discovered and some proposed modifications [7, 8, 9, 10, 11] are suggested in every methods to make it more effective.