语音增强系统中宽带噪声的滤除

针对宽带噪声的特点，探讨了利用谱减法的改进形式进行语音增强处理，以实验结果说明了参数ｍ、ｎ的取值与增强效果的关系，当ｍ、ｎ取值恰当，则可获得较为理想的增强效果。在估计噪声的短时能量谱时，采用了低信噪比自适应有声／无声算法来判别纯噪声的信号帧，实时估计噪声，提高了噪声估计的准确度，取得了较好的语音增强效果。

Suppression of Wideband Noise by Application of Speech Enhancement System

By recognizing the character of wideband noise, the method of improving spectral subtraction can be exercised to enhance the speech process. Experiments show that speech enhancement corresponds to good data fitting for the parameters m and n . The method of low signal to noise ratio correlated to making decisions for voiced or unvoiced actions relies on recognitionof the phase of pure noise signal and the real time noise estimate. All of the them are based on the short time spectral amplitude estimation.