Representation of Multiple Acoustic Sources in a Virtual Image of the Field of Audition from Binaural Synthetic Aperture Processing as the Head is Turned

摘要

The representation of multiple acoustic sources in a virtual image of the field of audition based on binaural synthetic-aperture computation (SAC) is described through use of simulated inter-aural time delay (ITD) data. Directions to the acoustic sources may be extracted from the image. ITDs for multiple acoustic sources at an effective instant in time are implied for example by multiple peaks in the coefficients of a short-time base (≈2.25 ms for an antennae separation of 0.15 m) cross correlation function (CCF) of acoustic signals received at the antennae. The CCF coefficients for such peaks at the time delays measured for a given orientation of the head are then distended over lambda circles in a short-time base instantaneous acoustic image of the field of audition. Numerous successive short-time base images of the field of audition generated as the head is turned are integrated into a mid-time base (up to say 0.5 s) acoustic image of the field of audition. This integration as the head turns constitutes a SAC. The intersections of many lambda circles at points in the SAC acoustic image generate maxima in the integrated CCF coefficient values recorded in the image. The positions of the maxima represent the directions to acoustic sources. The locations of acoustic sources so derived provide input for a process managing the long-time base (10s of seconds) acoustic image of the field of audition representing the robot’s persistent acoustic environmental world view. The virtual images could optionally be displayed on monitors external to the robot to assist system debugging and inspire ongoing development.