基于互相关函数的孤立词端点检测算法

提出了基于互相关函数的检测元音的端点检测算法。在语音识别系统中产生错误识别的原因之一是端点检测有误差,在高信噪比情况下,准确地确定语音的端点并不困难,然而在低信噪比情况下,一些常规的端点检测方法,例如基于能量的端点检测方法就不能有效地工作。本文利用互相关函数检测元音的端点,然后根据元音和辅音的能量比一般相对稳定的特点确定出实际语音端点。实验结果表明,本方法可以在不同信噪比下提高端点检测的准确率。

A Robust Endpoint Detection Algorithm Based on Cross Correlation Function in Isolated Word Recognition

An algorithm for endpoint detection which based on cross correlation function is presented. A major cause of error in auto- matic speech recognition (ASR) systems is the inaccurate detection of the beginning and ending boundaries of test and reference pat- terns. Accurate determination of endpoints of speech is not very difficult if the SNR is high. Unfortunately, the conventional speech detection methods such as energy cannot work well in noisy environments. In the paper, cross correlation function is used to detect the endpoint of the vowels. The energy ratio of vowels and consonants is relatively steady and accordingly the endpoint of actual speech can be detected. The experimental results show the performance of endpoint detection can be improved by this method in var- ious SNR environments.