低信噪比语音信号端点检测和自适应滤波

Low SNR Speech Signal Endpoints Detection and Adaptive Filtering

本文采用时间序列信号短时分形的非线性分析方法 ,对信噪比在 0dB和 -5dB的噪声语音信号进行了语音端点的检测和自适应滤波研究。通过计算机仿真实验 ,结果表明 :噪声语音信号的短时分形维维数的大小能较准确地判定语音的起止端点 。

This paper introduces the detection of noise-speech endpoints(i.e.beginning and end of noise-speech)and filtering on low SNR speech signal through application of time seguence short time fraction dimension.The simulation experiment by means of computer shows the short-time fraction dimension of signals can be used to accurately detect the beginning and endpoints of speech on the 0dB and -5dB noise speech signals.The adapting filter based on the short-time fraction dimension of signals can filter the noise of speech signal.