Tracking drones with small microphone arrays

Abstract

This report investigates how accurately drones can be tracked by measuring the noise they emit over time with a small microphone array; the ReSpeaker Mic Array v2.0. This acoustic localization is achieved by applying beamforming to select frequency ranges. In literature this localization has been achieved and accurately so, but the microphone arrays there are generally at least a meter in diameter with many microphones. The array used in this project is about 5 cm in diameter and contains only four microphones. It is therefore cheaper in cost and in terms of calculation time. The first experiment takes place in an anechoic chamber to compare the beamform output with reference values, specifying the expected elevation and azimuth angles. The output turns out to be fairly accurate, with an error of a couple of degrees. A second experiment is with indoor drone flights, where the reference location is only broadly known. The beamform output at high frequencies appears to be very accurate and the output at low frequencies is less accurate. This is expected due to the arrays small size which hinders beamforming at low frequencies but it is unfortunate as drones emit most of their noise at lower frequencies. The third experiment is with outdoor drone flights in the Unmanned Valley near Katwijk aan Zee. A 500 gram heavy Parrot Bebop 2 drone with three-bladed propellers can be localized up to a distance of 100m and a 907 gram heavy DJI Mavic 2 drone with two-bladed propellers can be localized up to a distance of 80m. Both with an error between GPS and beamform output of less than 10 degrees if beamforming is applied to a higher frequency range of 2-5kHz. This only applies if the array is placed at a 45-degree angle with the ground, which improves the beamforming accuracy. The application of functional beamforming did not improve the results of this experiment.