Over-determined Speech Source Separation and Dereverberation

In this paper, we propose a joint speech source separation and dereverberation technique which works well when the number of microphones is more than the number of speech sources. Microphones that exceed the number of sound sources are utilized for background noise reduction. The proposed method extends the recently proposed ILRMA-T into an over-determined technique. We reveal that an orthogonal constraint enables efficient update of a noise reduction filter in the proposed framework similar to the previously proposed over-determined speech source separation case. Secondly, the proposed method utilizes a joint diagonalization framework to reduce the residual noise signal in the output separated signal. Experimental results show that the proposed method efficiently separates speech sources in reverberant and noisy environments.