Underdetermined Blind Source Separation in Reverberant Environment

This work studies a blind source separation method for convolutive mixtures in an underdetermined case. Due to strong reverberation, the mixing model still presents convolution property even in the time-frequency (TF) domain. Based on the convolutive model in TF domain, a blind source separation is developed, and the resulting optimization problem is solved via an alternating framework where the alternating direction method of multipliers (ADMM) is developed. Experimental results show that the proposed method outperforms other methods in strong reverberant environment.