On the denoising of non-Gaussian noise in the acoustic array measurement

The microphone array can simultaneously obtain the space-time-frequency information of the sound sources, which is a foundational tool in acoustic imaging. There are mainly two categories of methods that have been proposed in the past. One is the Nearfield acoustical holography (NAH), and the other is the acoustic beamforming. However, most of the application of the two methods mentioned above are based on the Gaussian noise assumption, which is not always accurate in the on-site measurement. For example, the shock noise with skewed probability density function (PDF) may appear on the signal record when the turbulent eddies are not very controlled. Thus, more sophisticated noise should be considered when moving the measurement from the laboratory to the on-site environment. In this paper, the non-Gaussian noise is assumed in the array model, and it is modeled with the Gaussian mixture model. The signal from the sources of interest is finally recovered by the Expectation-maximization algorithm, which is an iteration between the low-rank approximation of the sound sources and the estimation of the parameter of the Gaussian mixture model. The proposed method is investigated with the simulation and validated with the experimental data.