Binaural cues characterization

« back to Robot Audition

Almost everyone agree on the physical meaning of interaural cues used for horizontal sound source localization (time delay/phase difference and amplitude difference between the left and right ears signals), there is no clear agreement on the best way to estimate their value. For instance, here is short, clearly incomplete, list of the more classical estimation methods:

The sensorimotor version of perception

We then worked during K. Youssef PhD on the comparison between the different definitions proposed in the literature, and on the evaluation of their performances evaluated through a positional classes separation criterion (in azimuth and distance), or in terms of speaker recognition rate. These works demonstrate the robustness of amplitude-bases cues w.r.t. reverberation (ands this was of course expected!), and show that cochlear frequency decomposition outperform classical FFT-based approaches, at a time where they were the most exploited in a Robotics context.

Binaural sound source localization >>