HUMANOID BINAURAL HEARING: ADAPTIVE APPROACH

摘要

When placed in a free sound field, a listener will obstruct an incoming sound wave, and ears, head and body will cause a linear filtering of the sound signal. This filtering is completely and uniquely described by the Head-Related Transfer Function (HRTF). In the general definition of this function all linear properties of the sound transmission are included. All proposed descriptors of localization cues, such as inter-aural differences in arrival-time, in phase, in sound level, as well as monaural cues, are embedded in the HRTF. They can thus be derived from it, whereas the opposite is not generally the case. Motivated by the rich content of the HRTF and by the role of the outer ear to direct, focus, and amplify sound, we present a new binaural method for sound source localization in 3D to be deployed for humanoid binaural hearing. Based on the content of the collected sound signals, the distance between the microphones reconfigures automatically in order to optimize the localization accuracy. Compared to other localization algorithms, the proposed system is outperforming in terms of localization precision and processing power.