Robust direction-of-arrival by matched-lags, applied to gunshots

Abstract

This work investigates the direction-of-arrival problem. A time-delay-estimate (TDE) obtained from a peak of a correlation function is subject to two types of error: type I, approximation errors, and type II, errors due to spurious signals. The iterative least-squares algorithm tentatively selects spatially coherent subsets of TDEs containing no type II errors and minor contributions of type I errors (“matched-lags”). Simulations use a seven-microphone array and a gunshot signal. The evaluation methodology is rigorous, comparing empirical distribution functions of estimation error of algorithms through two-sample, one-sided Kolmogorov–Smirnov tests, and quantifying differences with Cohen's D. The direction-of-arrival estimate is improved, specifically at low signal-to-noise ratios.