基于Teager能量算子的语音端点检测算法研究

Speech endpoint detection algorithm analyses based on teager energy operator

语音识别中端点检测是很重要的环节,检测的好坏直接影响到后面的语音识别的效果。传统使用的短时能量与短时过零率方法在信噪比较低时,不能有效地检测语音端点,检测准确率较低。利用Teager能量算子的非线性特性,能在抑制背景噪声的同时对平稳和不平稳信号有不同程度的衰减。因此,文中提出一种基于Teager能量算子的端点检测方法,并进行改进检测算法。经过实验证明,改进的算法与短时能量检测的结果相比,该算法在信噪比较低的情况下,能够比较准确地检测出语音的起始端点,同时语音端点检测准确率比较高,验证了该算法的有效性。

Because endpoint detection is a very important part in speech recognition,the effect of the detection influences the speech recognition directly. T he traditional methods that the short time energy and zero crossing rate can’t detect the speech endpoint effectively and have low accuracy when it has low SNR . The nonlinear characteristics of Teager energy operator can suppress the background noise,and have a different degree of attenuation to the stable and unstable signal at the same time. Therefore,this paper proposes a method of endpoint detection based on Teager energy operator, and improves the detection algorithm. After the experiment,the improved algorithm is compared with the results of short term energy detection. In the case of low SNR,the algorithm can accurately detect the start and end of the speech, and the accuracy of speech endpoint detection is relatively high,so the effectiveness of the proposed algorithm is proved.