The optimization of Speech Endpoint Confirmation research

Abstract

In speech processing, Speech Endpoint Confirmation has a very important position. Sub-band average-energy and the variance of the wavelet coefficients are the common algorithms, the disadvantage of the former algorithm is the limitation of only suitable for Gaussian white noise environment, while the latter algorithm can't overcome the drawback in Time-consuming, bad real-time. According to these problems, this paper puts forward optimization algorithm combined with the two methods. It improves applicability in speech endpoint confirmation. The MATLAB simulation results show that the system using the optimization algorithm achieved optimal accuracy and quickness, and gained anticipative effect.