针对谱减法用于语音增强残留噪声的问题,提出了一种新的阈值确定方法对小波系数进行分层阈值处理,并用短时能量和短时过零率对阈值处理过的各小波系数进行端点检测,利用检测结果确定语音信号的起止点并把起止点之间的部分取出形成新的小波系数向量,然后把经上述处理过的各层小波系数重构,以达到去掉噪声的目的。实验结果表明,与谱减法相比,文中方法可以明显提高语音信号信噪比和清晰度。 Aiming at the problem of spectral subtraction for speech enhancement residual noise, a new method of threshold determination is proposed to deal with the thresholds of wavelet coefficients. Thresholds of short-term energy and short-term zero- Detection, the use of test results to determine the start and end of the voice signal and the start and end points between the part of the removed to form a new wavelet coefficient vector, and then the above-processed wavelet coefficients reconstructed in order to achieve the purpose of removing noise. Experimental results show that compared with spectral subtraction, the proposed method can significantly improve the signal-to-noise ratio and sharpness of speech signal.