关键词: 语音端点检测, 小波分析, 模糊神经网络

Abstract: This paper presents a method for speech endpoint detection based on wavelet analysis and fuzzy neural network to improve the adaptability and robustness of speech endpoint detection. Firstly, the characteristic quantities of speech signals are obtained by the wavelet transformation; then the input to fuzzy neural network can be computed based on these characteristic quantities, and finally the signal’s type can be determined. This paper mainly introduces how to obtain the characteristic quantities of signals and how to establish the model and the learning algorithm of fuzzy neural network. The experiments show that this method has better adaptability and robustness and can detect signals with different SNR, compared with the traditional detection methods.

Key words: speech endpoint detection, wavelet analysis, fuzzy neural network