Binaural Sound Localization Based on Detection of Multi-band Zero-Crossing Points

A method of binaural sound localization based on detection of multi-band zero-crossing points generated from binaural filter-band outputs is presented. In the presented method, the ITD values are estimated by a histogram weighted by the estimated source noise rate(SNR) computed from the variances of ITD samples. As a result, the method provides an accurate estimate of sound source direction and is more robust to noise. To verify the validity of the method, simulations for sound localization are performed under the condition of white noise. The simulation results show that the method is able to provide an accurate estimate of sound source direction and is more robust to noise while offering significantly less computational complexity compared to the conventional cross-correlation based methods.