MFCC是语音识别中常用的特征参数,根据MFCC分量对语音端点的敏感性,提出利用平常舍去的识别特征参数分量MFCC0作为语音端点检测的参量。接着根据MFCC0的特性设计了一种新的端点检测方法,该方法简单且无需增加额外的计算量。实验结果表明,基于该方法的语音识别系统不仅可以通过端点检测大大压缩数据量,而且提高了系统的识别率。 MFCC is a commonly used feature parameter in speech recognition. According to the sensitivity of MFCC to speech endpoint, MFCC0 is proposed as the parameter of speech endpoint detection. Then based on the characteristics of MFCC0, a new endpoint detection method is designed. The method is simple and does not require additional computation. The experimental results show that the speech recognition system based on this method not only greatly reduces the amount of data compression through endpoint detection, but also improves the system recognition rate.