一种基于短时倒谱速变率的语音信号平滑端点检测方法

短波通信以其天波传播特性，在通信领域具有其他通信手段无法替代的地位。为解决短波通信中信噪比较低，噪声信号严重影响语音处理效果的问题，提出了一种基于短时倒谱速变率的平滑端点检测方法，通过检测噪声信号的倒谱特征，初步确定语音信号端点，然后加入平滑优化处理以及Holdon设计，降低由于倒谱突变造成的误判。实测信号仿真实验证明，该方法在不过多加重系统负担的前提下，取得了良好的效果。

Method of Speech Signal Smoothing Endpoint Detection Based on Short-term Cepstrum Quick-changing

Shortwave communication occupies an irreplaceable position in the communication domain because of its characteristic of sky wave propagation. A method of the speech signal smoothing endpoint detection based on short-term cepstrum acceleration is proposed to solve the problem of the low SNR in shortwave communication, which severely influences the result of speech psocessing. In this method, the preliminary determination of the speech signal endpoint is obtained by detecting the cepstrum characteristic of noise signal, and then smoothing optimization processing and Holdon design are adopted to reduce the erroneous judgement caused by the cepstrum quick changing. The simulation experiment of actual shortwave signal pro cessing shows that the above method gets a satisfactory result without putting over burden to systems.