机构地区: 湖南师范大学物理与信息科学学院
出 处: 《微计算机信息》 2007年第30期228-229,264,共3页
摘 要: 提出了一种新的基于小波和时频分解的语音端点检测方法。首先通过小波分解对含噪信号进行增强,然后采用Matching pursuits算法对去噪信号进行时频分解,使得信号在时频平面上具有较明显的魏格纳能量分布,最后利用该特点设定合适的门限来进行语音端点检测。实验结果表明,该方法对低信噪比的语音端点检测仍有效。 A new method of speech endpoint detection based on wavelet and decomposition in time-frequency domain is proposed. Signal containing noise is strengthened through wavelet decomposition, then the reconstructed signal is decomposed by Matching pursuits algorithm in time-frequency domain to obtain apparent Wigner energy distribution. Finally, a suitable threshold is selected to detect the speech endpoint according to these properties. Experimental results demonstrate this approach is effective for low SNR speech endpoint detection.
领 域: [自动化与计算机技术] [自动化与计算机技术]