关键词: 小波分析, 神经网络, 语音端点, 特征提取, 特征选择

Abstract: In order to improve the adaptability and robustness of speech endpoint detection, this paper proposes a speech endpoint detection method based on wavelet analysis and neural network（WA-PCA-RBF）. The features of speech signals are extracted by wavelet analysis; the features are selected by Principal Component Analysis to remove redundant features; the selected features are input to RBF neural network to build the speech endpoints detection model in which the RBF neural network’s parameters are optimized by genetic algorithm. The results show that the proposed method has improved the detection rate, and it has better adaptability and robustness in complicated noise environment compared with the traditional detection methods.

Key words: wavelet analysis, neural network, speech endpoints, feature extraction, feature selection