Probabilistic Spectrum Based Azimuth Estimation with a Binaural Robotic Bat Head

摘要

A small size in-air sonar system is described that performs binaural spectrum based localization. We show that, for a given receiver configuration, a signature of the target's azimuth can be extracted by combining the head related transfer function of the individual receivers. The posterior probability density function of the target's azimuth is calculated by comparing a binaural spectrum measurement with a set of binaural azimuth templates. We propose a Bayesian framework to combine multiple measurements, in particular measurements collected from different viewpoints and/or measurements collected with altered receiver configurations. We show that combining consecutive measurements from multiple viewpoints increases the reliability of the azimuth estimate, even at larger target elevations. An example of the combination of estimate results using two different receiver configurations is presented, showing the potential for actively reducing estimate ambiguities. Furthermore, we show how the antenna features of the artificial pinna shapes can increase bearing estimation performance.