摘要
在语音识别系统中,端点检测技术对于系统的识别准确率来说是至关重要的。提出一基于小波子带能量和小波系数方差的语音端点检测算法。和其他传统的端点检测方法如短时能量、过零率方法等相比,该算法更加有效。计算机仿真结果证明了该算法更适合于语音端点检测,尤其是在低信噪比(SNR)条件下。
In the process of speech recognition,accurate endpoint detection is crucial for good speech recognition accuracy.In this paper,we propose a novel algorthm based on wavelet sub-band energy and wavelet coefficient variance.In the comparison of the algorthm to the other traditional method such as energy,ZCR etc., the algorthm proved effective.The SA is shown to be well suited for the detection of speech endpoint,especially for lowsignal-to-noise ratio(SNR).
出处
《微型电脑应用》
2009年第11期36-37,5,共2页
Microcomputer Applications
