Generalized amplitude interpolation by β-divergence for virtual microphone array

In this paper, we present a generalization of the virtual microphone array we previously proposed to increase the microphone elements by nonlinear interpolation. In the previous work, we generated a virtual observation from two actual microphones by an interpolation in the logarithmic domain. This corresponds to a linear interpolation of the phase and the geometric mean of the amplitude. In this paper, we generalize this interpolation using a linear interpolation of the phase and a nonlinear interpolation of the amplitude with adjustable nonlinearity based on β-divergence. Improvement of the array signal processing performance is obtained by appropriate tuning of the parameter β We evaluate the improvement in speech enhancement using a maximum SNR beamformer.