Energy-Based Multiple Source Localization with Blinkies
Daiki Horiike, Robin Scheibler, Yuma Kinoshita, Yukoh Wakabayashi, Nobutaka Ono · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference · 2020
We propose energy-based multiple source localization using sound power sensors called Blinkies that we have recently developed. A Blinky consists of a microphone, LEDs, a microcontroller, and a battery. The intensity of the LED is varied by sound power. Namely, Blinkies work as sound-to-light conversion sensors. They are easy to distribute over a large area, and thus, sound power information can be obtained by capturing the Blinky signals with a video camera. When multiple sources are present, their sounds are mixed, and Blinky signals also reflect the power of the mixture. The idea of the proposed source localization is to decompose a multiple-sources localization problem into “single source localization” problems. More specifically, the Blinky signals can be factorized into transfer function gains and temporal activations by non-negative matrix factorization. Each obtained gain vector is used for estimating each source location. We conduct numerical simulations to evaluate the performance of this method in indoor space like a meeting room. The experimental results show that the proposed framework using Blinkies is effective.