多特征和支持向量机相结合的语音端点检测模型

Speech endpoint detection model based on multiple features and support vector machine

为提高语音端点检测正确率,提出一种基于多特征和支持向量机相结合的语音端点检测模型。首先提取多种语音特征,并将它们组合在一起,然后将组合特征输入到支持向量机训练建立相应的语音识别模型,最后采用建立模型对语音信号进行检测和识别。仿真结果表明,与其他检测模型相比,多特征融合和支持向量机的检测模型提高了语音端点检测正确率,具有更好的适应性和鲁棒性,对不同信噪比的信号都有较好的检测能力。

In order to improve the rate of speech endpoint detection, this paper presents a speech endpoint detection model based on multiple features and support vector machine. Firstly, the features of speech signals are obtained and combined together; then they are input to support vector machine to build the speech endpoint detection model, and finally the signal is detected by the built model. The simulation experiment results show that, compared with the other detection models, the proposed model has improved the detection rate, and has better adaptability and robustness; it can detect signals with dif- ferent signal to noise ratio.