A real-time blind source separation for speech signals based on the orthogonalization of the joint distribution of the observed signals

This paper proposes a real-time BSS (Blind Source Separation) method for acoustic signals. The proposed method can estimate the unknown source signals based on the orthogonalization of the joint distribution of the observed mixture signals. For a real-time separation, we calculate whitening, rotation, scaling adjustment and output channel selection for a short-time frame. From the simulation results, the proposed method can estimate the original source signals for not only a human utterance noise but also a stationary noise.