Robust DOA estimation of heavily noisy gunshot signals

Direction of Arrival (DOA) estimation of gunshots is an important asset to law-enforcement agencies and defense forces, for shooter localization is key to improving public and troop safety. Solutions vary according to scenario, application, and available resources. As the distance between the firing position and the sensor array increases, as in a typical sniper scenario, signal to noise ratio decreases and the estimation degrades. This paper proposes a gunshot DoA estimation algorithm to be used with highly noisy signals. We combine exhaustive search for selecting pairs of microphones from the array to attain the best DOA estimation results and fast response time for different shooting scenarios. We are particularly interested in highly corrupted signals for which state-of-the-art algorithms fail. Experimental results from simulated and recorded gunshot signals were used to evaluate the performance of the proposed scheme.