SENSOR ARRAY-PROCESSING PERFORMANCE EVALUATION VIA BOOTSTRAP RESAMPLING

This paper examines bootstrap resampling methods for evaluating the pe rformance of source localization methods on real sensor array data wit hout precise a priori knowledge of true source positions and the under lying data distribution. Bootstrap resampling methods are used here to measure the standard deviation of bearing estimates achieved by perfo rming delay-and-sum, minimum variance, and MUSIC spatial spectral esti mation on a real narrow-band towed-array data set. Some simple theoret ical guidelines are given to indicate the sample size required to achi eve accurate performance estimation. Results suggest that quantitative bearing estimation performance comparisons can be made within the obs ervation time constraints imposed by source and field dynamics. Furthe r, comparisons of bootstrap estimates to Monte Carlo and theoretical b enchmarks provide a means of validating the parametric models assumed by these conventional analysis techniques.