基于小波阈值去噪的麦克风阵列语音增强方法

现有的麦克风阵列语音增强方法中的延迟-求和波束形成算法只对不相干噪声或弱相干噪声有一定的消噪能力，如果语音中混有较强相干噪声，则此传统的方法对其没有消除能力。针对这个局限性，文中把小波阈值去噪的方法与传统的延迟-求和波束形成算法有效结合，使其对相干噪声也具有很好的消噪能力，同时减少由于噪声的存在而引起的时延估计误差，提高时延估计的准确性，使最终求和结果更好。通过仿真结果表明，这种改进方法可以改善最终语音效果，提高语音清晰度，使人耳更好地接受。

Microphone Array Speech Enhancement Algorithm Based on Wavelet Threshold Denoising

Delay-sum beamforming method as a kind of microphone array speech enhancement algorithms only is effective for irrelevant noise and weak coherent noise. If the speech mixes coherent noise,the conventional method cannot remove it. For this limitation,in this paper wavelet threshold denoising is applied into conventional delay-sum beamforming speech enhancement algorithm,making it to be ef-fective for removing coherent noise. While reducing the time estimation error caused by the noise,improve the accuracy of time delay esti-mation and make the final sum result better. The simulation results show that the improved method is effective,which improves the charity of speech and degree of ear acceptance.