Robotic Sound Source Localization Using Bio-Inspired Acoustic Sensors

In sound source localization, there is a fundamental size limit; the smaller the size, the smaller the directional cues that are relied on to pinpoint the sound source. As such, it is challenging to develop miniature sound source localization robotic system where space is too confined to employ conventional microphone arrays without compromising localization performance. Our previous studies show that through mechanical coupling with well-tuned structural parameters, directional microphones inspired by the parasitic fly Ormia ochracea can amplify the minute interaural time delay (ITD) by more than ten times, which enables the reduction of device size significantly while maintaining localization performance. In this paper, Cramer Rao lower bound (CRLB) is derived for the fly-ear inspired sensor and the conventional directional microphones to study the effects of mechanical coupling on the decrease of the theoretical lower bound of azimuth estimation. This improvement gives mobile robots the capability to reactively localize sound in an indoor environment. Using this miniature sensor, new sound source localization method is proposed to localize a stationary sound source in 2-D (azimuth and elevation). In the proposed sound localization method, Model-Free Gradient Descent (MFGD) optimization method, one of the main challenges is to choose the appropriate cost function to achieve minimum number of iterations and the smallest absolute error. To this end, different cost functions are proposed and investigated with different control schemes. Simulation results showed the ability of this technique to solve the ambiguity problem and localize the sound source.