Localization of multiple simultaneous speakers by combining the information from different subbands

Time Difference Of Arrival (TDOA)-based algorithms are the main methods for speech source localization. A category of these methods are based on Generalized Cross Correlation (GCC). These methods estimate the source location based on the calculated TDOA between microphones signals. The accuracy of these methods decreases as the amount of noise and reverberation increases. In this paper, we propose the utilization of subband processing for the localization of two simultaneous speech sources. While the conventional methods consider the whole signal spectrum identically in the localization procedure, the proposed method takes advantage of the differences in the frequency bands of the mixed speech for the localization of multiple speakers. Actually, the proposed method computes the GCC in the different frequency bands and then, combines the information from the subbands in a so-called smart manner. We have discussed several approaches for the combination of subband. The performance evaluations in different environmental conditions demonstrate the superiority of the proposed method compared to the fullband GCC method. The proposed method considerably increases the accuracy of simultaneous speaker localization.