【摘要】 语音信号与随机噪声在不同尺度上进行小波变换时 ,其小波变换系数和尺度大小的特性关系存在着不同的特征表现 ,而且 ,浊音和清音也各有其特性 .鉴于此特点 ,给出了一种基于小波变换的多尺度多阈值语音增强方法 .该方法采用软限幅函数对浊音和清音信号的小波变换系数作不同的阈值处理 ,既抑制了噪声 ,又减少了语音段的信息的损失 ,提高了信噪比 .仿真结果说明 ,这是一种有效的语音增强方法 .更多还原

【Abstract】 There are different characteristics in the relationship of wavelet transform coefficient and scale when the speech signal and the random noise are transformed on the different scale,and noisy sounds and unvoiced sounds are also characterized with different features. According to this,the paper gives a speech enhancement method of mult scale and mult threshold based on wavelet transform. This method uses soft threshold deals wavelet transform coefficient of noisy sounds and unvoiced sounds with different threshold. It restrains noise and reduces the loss of speech region information, and enhances SNR. The simulated results shows that it is an effective speech enhanced method.更多还原