A Novel Sound Localization Method Based on Head Related Transfer Function

In machine sound localization, convenient and effective estimation modes are required. A dummy head with certain head related transfer function (HRTF) can be introduced to sound localization instead of the conventional microphone array or other sensor systems. Since the spectrum of HRTF varies with the sound source direction, a sound localization method based on spectral cues is presented. By simplifying the direction estimation process of human spatial hearing, the directional information is extracted from the spectrum of HRTF and is multi coded in parallel with a feed forward neural network. Then the multi coded directional information is weighted to estimate the sound position. Simulations show that the proposed method is effective.