Investigating the potential of UAV for gunshot DoA estimation and shooter localization

Abstract

This paper presents an approach to obtain stationary shooter localization in a Cartesian plane using a moving array of microphones. A single array is embedded in an unmanned aerial vehicle (UAV), a quadcopter, to explore benefits provided by its intrinsic mobility. We propose a model which is based on geometrical acoustics and that assumes the gunshot signals as being buried in strong noise generated by propellers. Generalized cross correlation algorithms are used to estimate the Direction of Arrival (DoA) of the impulsive gunshot signals and, finally, bearings-only target motion analysis techniques are applied to estimate potential shooters localization from the noisy DoAs.

Bearings-only aerial shooter localization using a microphone array mounted on a drone

“…This 2D setup was employed because the quadcopter was placed at the same height of the shooters implying in a fixed elevation angle close to zero (or, equivalently, a zenithal angle close to 90 degrees). Also, given the preliminary results presented in [4] pointing to the fact that signal enhancement schemes are not effective for all distances (between the array and the shooter) in this kind of application, we have processed the signals (detecting and estimating the DoAs) directly from raw data, without signal enhancement.…”

“…The Exhaustive Search (ES) is another approach that selects a subset of n pairs of microphones: in this case, it evaluates the least-squares cost function yielded by all possible n cross-correlations and chooses the combination that corresponds to the least one [17]. Due to previous results on DoA estimation of gunshot signals collected by a quadcopter [4], the exhaustive search algorithm is used in this work.…”

“…Therefore, we shall use TLS algorithm [18] for shooter position estimation from gunshot bearings obtained from field recordings. This method was also chosen because of its superior performance over the Linear Least-Squares algorithm which performance has been compared in [4].…”

“…Although effective in terms of mitigating the multipath propagation problem, the aerial balloon based solution does not meet today's challenges when it comes to cost efficiency and mobility. Owing to recent advances in modern technology, drone based solutions as the one proposed in [4] have become a feasible option and can be produced in large scale at a relatively low cost.…”

“…Therefore, the use of signal enhancement techniques such as adaptive filtering, spectral subtraction, and median filtering [5], [6], [7], [8], may be needed to improve the SNR prior to applying the algorithms for DoA estimation. A preliminary discussion on this issue has been presented in [4]. However, having an estimate of the DoA of the muzzle blast and of the shockwave may not be enough to pinpoint shooter's location.…”