A hough transform based approach to acoustic shot localization
Yifu Luo, János Sallai, Ákos Lédeczi · 2016
The paper presents a novel sensor fusion technique to acoustic multi-shooter localization using a network of sensors detecting the time and angle of arrival of high-energy acoustic events. The unique challenge is that the sensors do not to classify the acoustic signal as shockwave or muzzle blast, and would trigger on non-gunshot related noise, as well. While this allows for low cost and power-efficient hardware designs, the computational burden of classifying sensory data, rejecting false positive detections, as well as associating sensor detections with multiple simultaneous gunshots events, is shifted to the sensor fusion implementation. The novelty of this work is that it relies on an innovative use of the Hough transform (HT), an image processing algorithm that locates shapes in raster images, to solve an acoustic source localization problem. The first contribution of the work is a proof-of-concept HT based approach that, given a set of timestamped angle of arrival detections, associates detections with wavefronts of acoustic shockwaves, and estimates the corresponding bullet trajectories. The second contribution is an algorithm that associates detections with muzzle blast wavefronts, and estimates the corresponding shooter positions. Preliminary results on the performance of data association, as well as on the accuracy of trajectory and shooter position estimation are presented using synthetic data.