Dynamic Localization of Ambiguous Sounds by Drones

Abstract

The ability to locate sound sources from a drone is crucially important in many applications. Drones with such ability can be used in search and rescue (SAR) missions to rapidly locate distress sound signals of victims and rescuers. An array of spatially separated microphones can be used on drones to compute the time difference of arrival (TDOA) of broadband sounds and to estimate the sound direction. However, for narrowband sounds, due to the periodicity in the sequence of sound samples, there is ambiguity in TDOA estimations and it is impossible to identify the true direction of the sound. This work describes a novel method for localizing ambiguous narrowband sounds, such as the sound of safety whistles or personal alarms in SAR operation, from a rotorcraft drone. This is achieved by employing dynamic cues resulting from in-place platform rotations of the drone and by fusing acoustics and attitude information together.