为提高低信噪比下语音端点检测的准确性,提出了一种基于模糊熵与改进相关向量机的端点检测算法.首先对语音信号进行基于听觉感知特性的语音增强,然后提取每帧信号的模糊熵作为改进相关向量机的输入矢量,同时针对单一核函数对预测分类鲁棒性弱的问题,对不同核函数进行自适应多核组合,融合多个核函数的特性,提高分类精度和鲁棒性.实验结果表明:在低信噪比环境下,基于模糊熵与改进相关向量机的端点检测能更有效地检测出语音的端点,准确率达到93.2%. To improve the accuracy of voice endpoint detection at low signal-to-noise ratio (SNR), an endpoint detection algorithm based on fuzzy entropy and improved correlation vector machine is proposed.Firstly, speech enhancement based on auditory perception is performed on speech signal, and then the signal of each frame is extracted Fuzzy entropy is used as the input vector to improve the correlation vector machine. At the same time, aiming at the weak robustness of the single kernel to the prediction classification, the adaptive multi-kernel combination of different kernel functions and the fusion of multiple kernel functions are proposed to improve the classification accuracy and robustness The experimental results show that endpoint detection based on fuzzy entropy and improved correlation vector machine can detect the endpoints of speech more effectively, with an accuracy of 93.2% under low SNR environment.