Indoor multiple sound source localization using a novel data selection scheme

The multiple sound source localization problem in reverberant environments is studied by using a circular microphone array. The problem is approached by performing single source localization at each of the selected time-frequency (TF) data points from the received signals followed by a histogram technique to achieve robust localization. A novel TF data point selection scheme is proposed based on the ratio of the power of a TF point to that of its adjacent preceding TF point at the same frequency. The method introduces little extra computation and is suitable for real-time implementation. The proposed algorithm is compared with the algorithm in [1] to show its effectiveness and efficiency using extensive simulations.