一种低信噪比下语音端点检测的改进算法

An Improved Algorithm for Speech Endpoint Detection in Low SNR Environment

端点检测在语音信号的处理与识别过程中占据着极为重要的位置。为了解决其在低信噪比环境下检测准确率低的难题,提出一种改进的语音端点检测方法。该方法首先使用Boll的改进谱减法对带噪语音降噪,然后求出降噪后每帧语音的对数能量与自相关函数余弦角值之比,最后把对数能量-自相关夹角余弦比值用作双门限算法的参数,用于语音段起止点位置的判断。经仿真实验验证,与其他传统端点检测算法相比,该算法在低信噪比环境中具有更高的端点检测准确率,对语音信号的后续处理也提供了较大帮助。

Endpoint detection occupies an indispensable position in the process of speech signal processing and recognition.In order to solve the problem of low detection accuracy in low signal-to-noise ratio environment,this paper proposes a speech endpoint detection algorithm based on improved spectral subtraction,which combines logarithmic energy and cosine angle of autocorrelation function.In this method,Boll's improved spectral subtraction method is first utilized to reduce the noise of noisy speech,and then the ratio of logarithmic energy and cosine angle of autocorrelation function of each frame of speech after denoising is obtained.Finally,this value is used as the parameter of the double threshold algorithm to judge the starting and ending position of speech.The simulation results confirm that compared with other common endpoint detection algorithms,this algorithm has higher detection accuracy in low SNR environment and is of great help to the subsequent processing of speech signals.