Binaural source localization using a HRTF data model with enhanced frequency diversity

A novel method for binaural source localization using a HRTF data model with enhanced frequency diversity is proposed in this work. The method proposed herein is developed in the frequency domain and enhances the directional information in the sound signal for efficient broad band sound source localization even under conditions of low SNR. The directional information is enhanced by cancelling the spectrum of the direct sound source. Subsequently the sound source is localized using a sub space based method. Binaural source localization experiments are performed using HRIR from the CIPIC database and the HRTFs recorded from a customized mannequin developed using 3D printing technology. Experimental results on sound source localization indicate a reasonable improvement in terms of RMSE when compared to state of the art methods.