Sound Source Tracking by Drones with Microphone Arrays
Taiki Yamada, Katsutoshi Itoyama, Kenji Nishida, Kazuhiro Nakadai · 2020
This paper presents a method for sound source localization and tracking using drones with microphone arrays. Microphone array processing is an established technique for estimating the direction of a sound source. However, for search and rescue applications, mere sound source direction is insufficient since sound source location in terms of direction and distance is essential. Accordingly, we propose a novel method we call Multiple Triangulation and Gaussian Sum Filter Tracking (MT-GSFT). MT-GSFT obtains the sound source location by triangulation using microphone arrays attached to multiple drones. Since triangulation results differ among pairs of arrays, they are transformed into a probabilistic representation using a Gaussian mixture, then a Gaussian sum filter is applied to estimate the sound source trajectory. Numerical simulation demonstrated MT-GSFT is effective in removing outliers in sound source localization to track sound sources better than the Linear Intersection method.