Sound Source Localization Using Head-Related Transfer Functions and Weighted Error Function

This study proposes a sound source localization method using binaural input signals. The method is based on the head-related transfer function (HRTF) database and the interaural transfer function (ITF) obtained from two measured input signals. An algorithm to reduce the effect of background noise on the localization performance in a noisy environment was adopted in the proposed localization method. Weighted error functions (WEFs), defined using the ITF and the ratio of HRTFs for two ears, were used with a special frequency weighting function derived to reduce the effect of noise and to render the WEF a hysical meaning. Computer simulations confirmed that the weighting function can effectively reduce the effect of background noise on the localization performance even if the noise power is very high. Localization tests in an actual room confirmed that both the azimuth and elevation angles of sound source can be estimated simultaneously with high accuracy. In particular, the front-back and up-down confusions, which are critical limitations for conventional localization methods, could be resolved using two input signals.