An Efficient Multi-Source DOA Estimation Algorithm for Underdetermined System

We propose a two-stage algorithm for the estimation of direction-of-arrival (DOA) of speech signals in a multi-source environment under the assumption that the source signals are sparse in the time-frequency (TF) domain. The signals are first separated, in each frequency bin, from their mixtures via time-frequency masking. The generalized cross-correlation method is subsequently used in the second stage to simultaneously estimate the source directions. The proposed algorithm does not require one to solve the permutation problem and is suitable even for under-determined cases where the number of source signals is greater than the number of sensors.