Speaker localization and speech separation using Phase Difference versus Frequency distribution

This paper proposes a novel sparse source separation method using a pair of microphones. The method is based on time-frequency (T-F) decomposition, applies the weighted Hough transform to the Phase Difference (PD) versus Frequency (PD-F) distribution of received mixture signals, and estimates source directions. Then, the estimated source directions and harmonic structure are used to separate the mixture signals. The effectiveness of the proposed method is shown through experiments in real acoustic circumstances.