Abstract—This paper proposes a blind source separation method using time-frequency (T-F) decomposition and clustering for estimated propagation direction vectors at T-F slots. The method is applicable to arbitrary array configuration in 3-D space and even in the presence of spatial aliasing. To solve spatial aliasing problem which causes ambiguity of phase, unity norm property of propagation direction vector is employed for solving phase difference ambiguity. With combining our previous direction-of-arrival (DOA) estimation algorithm and clustering in terms of spatial information, efficient separation procedure is achieved by a binary masking in T-F domain. Experimental results demonstrate that the proposed procedure effectively separate three or four speech sources with tetrahedron microphone array with wide sensor spacing where spatial aliasing may occur. 
 

Index Terms—microphone array, source separation, time-frequency masking, DOA estimation, spatial aliasing, kernel density estimation

Cite: Masashi Sekikawa and Nozomu Hamada, "Source Separation for Arbitrary Array Configuration in the Presence of Spatial Aliasing," International Journal of Signal Processing Systems, Vol. 4, No. 1, pp. 32-36, February 2016. doi: 10.12720/ijsps.4.1.32-36