Denoising method of partial coherent noise of microphone array based on Opt-Shrink

The denoising of microphone array signal is of great significance to its beamforming. In complex interference environment, the distribution of background noise no longer satisfies the traditional hypothesis of mutual incoherence, but tends to be partial coherent. Therefore, the distribution mechanism of spatial noise and the theory of partially coherent noise are studied, and a denoising method for partially coherent noise of microphone array under the condition of known number of sound sources is proposed. Based on the low rank assumption of sound source and the sparse assumption of partially coherent noise, the low rank part of cross spectrum matrix of microphone array is extracted iteratively based on the Opt-Shrink method. The simulation results show that the method can obtain unambiguous image results when the number of coherent channels is 10 and 25. However, the traditional diagonal removal (DR) method for incoherent denoising has poor sound source localization results for the large number of coherent channels. In the experiments of sound source localization under strong interference and low signal-to-noise ratio, the proposed method can get better denoising effect than the diagonal removal method.