摘要
提出了一种基于小波阈值去噪的语音增强算法。根据含噪语音中噪声、清音和浊音的不同特点,首先对它们进行检测区分,然后采用改进的Garrote阈值函数分别对噪声、清音和浊音运用不同的阈值方案进行处理。该方法在很大程度上抑制了噪声,又减少了语音段清音的损失,提高了信噪比,在MATLAB7.1中的仿真实验结果表明,该方法具有较好的增强效果。
This paper proposed a speech enhancement algorithm based on wavelet threshold denoising. According to difference of charac- teristics of Noisy Speech with noise, the unvoiced and voiced, firstly they were measured and distinguished, then the unvoiced, voiced and noise are processed by the improvement of the threshold Garrote function. To a large extent, this method suppressed noise, but also re- duced the loss of the unvoiced of the speech segment, especially in the high signal to noise ratio, improved signal to noise ratio. In Mat- lab7.1 the simulation results showed that the method has a good performance.
出处
《软件导刊》
2010年第2期43-45,共3页
Software Guide
