藏语语音端点检测方法研究

Study on Detection Method of Tibetan Speech Endpoint

语音处理技术包括很多环节,而端点检测是其中必不可少的环节,检测结果将直接影响到系统的性能。端点检测的目的是在噪声存在的情况下确定语音信号的起止点,属于语音信号处理的前端操作,为后续的一系列操作作充分准备。语音检测效果受环境影响,低噪音状态下,语音端点的检测结果就不太显著,为了解决这个问题,将谱减法和能零比法相结合。借助谱减法提高语音信号的信噪比,通过能零比法对处理后的语音进行端点检测,最后还需要对藏语语音进行实验分析。

Speech processing technology includes many links,and endpoint detection is one of the essential links,the detection results will directly affect the performance of the system.The purpose of endpoint detection is to determine the start and end point of speech signal in the presence of noise,which belongs to the front-end operation of speech signal processing,and makes full preparation for a series of subsequent operations.The effect of speech detection is affected by the environment.In the low noise state,the detection result of speech endpoint is not very significant.In order to solve this problem,spectral subtraction and energy zero ratio method are combined.With the help of spectral subtraction to improve the signal-to-noise ratio of speech signal,the end-point detection of the processed speech is carried out through the method of energy zero ratio.Finally,the experimental analysis of Tibetan speech is needed.