多种噪声环境下语音检测算法

Voice detection algorithm of various noisy

提出针对语音检测的一种新的算法-傅立叶谱矩阵,该方法以傅立叶变换为基础,对含噪语音信号直接进行分帧,对每帧做傅立叶变换后以帧为单位,构建能反映语音特征频段及能量的傅立叶谱矩阵。在傅立叶谱矩阵中找到语音信号能量的集中区域,从而找出所需的语音特征谱矩阵的范围,达到对语音的有效检测。

The paper refers to the new voice detection algorithm--Fourier Spectrum Matrix, which is based on Fourier Transform. Then the noisy voice signal is divided into frames with each of the frames for Fou- rier Transform, then in a frame unit, construct Fourier Spectrum Matrix which can reflect the voice＇s fea- ture frequency range and energy. In the Fourier Spectrum Matrix we find the voice signal energy concen- trated area, so as to find the extent of speech feature spectrum matrix and to achieve effective detection.