A combined approach using subspace and beamforming methods for time-frequency domain blind source separation

Blind source separation (BSS) is an approach to estimate original source signals only by measurements through multiple sensors without any prior knowledge about their mixture process. BSS is currently applied to a broad range of fields, such as pre-processing of speech recognition, analysis of bio-signals, etc. In BSS of convolutive mixtures, the signals must be expanded into time-frequency domains. In this case, permutation between frequencies becomes indefinite. This paper proposes a method for solving the indefiniteness problem by using directional peaks in addition to directional nulls in directivity patterns. This method makes it possible to combine the beamforming and subspace methods and to improve the accuracy in separatability of BSS. The effectiveness of the proposed method is demonstrated by computer simulation.