GCC-based DoA estimation of overlapping muzzleblast and shockwave components of gunshot signals

The problem of sniper detection is considered in this work. Audio signals recorded from gunshots by a microphone array are used for estimation of Direction of Arrival of muzzleblast and shockwave components; nevertheless, eventually, there may occur an overlapping of these signals. Algorithms from the Generalized Cross-Correlation (GCC) family are utilized in the DoA estimation, and it is shown that the time-lags that give the two largest peaks in the correlation functions correspond to muzzleblast and shockwave alignments. Although the Phase Transform method gives particularly good results in separating the peaks, a muzzleblast DoA estimation based on the maximum correlation peak obtained by the other GCC techniques are in general more accurate.