一种改进的多特征值语音端点检测方法研究  

An Improved Method of Multiple Eigenvalues for Endpoint Detection Research

作　　者：李申[1] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《电声技术》2015年第6期45-48,77,共5页Audio Engineering

摘　　要：为了提高语音识别系统中语音端点检测的准确性,提出一种改进的多特征值语音端点检测方法。新方法首先对信号进行小波分解及小波去噪处理,对去噪后小波子带系数进行短时能量与基音周期两特征值的提取,综合考虑两特征值的大小来进行语音端点检测。实验证明,改进的检测方法提高了端点检测的抗噪性及准确度。In order to improve the accuracy of speech recognition system of speech endpoint detection, this paper puts for- ward an improved method of multiple eigenvalues voice signal endpoint detection. The new method firstly precesses the signal via wavelet decomposition and wavelet denoising, and then extracts two eigenvalues, short -time energy and the pitch period for the wavelet subband coefficients after denoising. Lastly, comprehensively considering the values of two eigenvalues for voice signal, the speech endpoint could be detected accurately. The simulation results show that the presented improved detection method can improve anti - noise property and accuracy of endpoint detection.

关 键 词：端点检测 小波阈值去噪 短时能量 基音周期 

分 类 号：TN912.34[电子电信—通信与信息系统]