A Comparative Analysis on Speech Enhancement and Coding Techniques

Abstract

The speech data is usually a degraded speech data, which is a combination of clean speech and various noises like vocal, animal noise, instrumental noise and speech from other speakers. The noisy speech data collected and studied in detail for various kinds of noises. The spectral subtraction one of the best algorithms proposed for the enhancement of single channel speech. The speech enhancement is based on two aspects i.e., Noise estimation and Speech estimation, where noise estimation is critical part which has major impact on the quality of enhanced speech. Speech enhancement is used for many applications such as mobile phones, communication, speech recognition etc. The aim of this paper is to provide a comparison study of the different forms of speech enhancement techniques and algorithms by using filters like Kalman filter and Weiner filter on the output of SS- VAD which gives improved enhanced speech signals. Coding techniques like Code Excited Linear Predictive (CELP) Coding and Algebraic Code Excited Linear Predictive (ACELP) Coding algorithm, is also discussed the performance of these are compared with LPC algorithm for speech enhancement.