Learning sound location from a single microphone
ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation
Hi-index | 0.00 |
In this paper we combine monaural estimates of elevation with binaural estimates of azimuth and elevation. The monaural estimates are derived from the second difference of the amplitude spectrum, and are useful in the median plane, where binaural elevation estimates are very poor; however, they are source dependent, and do not provide azimuth information. The binaural estimates are derived from the interaural intensity difference, and are accurate away from the median plane. For a white-noise source in an anechoic environment, we show that the combination is accurate to within 2 degrees for azimuth and 12 degrees for elevation. Good performance is also obtained for speech, but accuracy degrades as the source bandwidth is reduced.