Performance analysis of capon matched-field algorithm in spatially correlated noise field

Matched field processing (MFP) for passive source localization is a comprehensive application of signal processing techniques in extremely complex environment. In essence, it is a parameter estimation processor by exploiting full wave modeling of acoustic waveguide propagation. Typical MFP performance demonstrates a threshold behavior, that is, below some signal-to-noise ratio (SNR), the mean-square error (MSE) increases dramatically. In this paper, we develop approaches to study MFP performance of the data-based Capon algorithm in the presence of spatially correlated noise. Especially, the effect of surface-generated noise (multi-rank) on MFP performance is interesting, different from that of point interference (one-rank). The Capon algorithm has a well-know high resolution. However, it cannot achieve the Cramer-Rao bound (CRB), because of the biased statistic and the SNR decrease due to sample covariance matrix. In this thesis, we modify the CRB to redefine the asymptotic local bound of the Capon algorithm here, and develop a so-called method of interval errors (MIE) to characterize the corresponding global performance bound. Performance analysis of capon MFP for source localization is implemented in a realistic shallow water environment. Simulation results suggest that, 1) in the presence of surface-generated noise or a point interference, both the threshold SNRs increase a lot, compared with the only white noise case; 2) surface-generated noise is a stronger destroyer of MFP performance than both white noise and a point interference. Nearly the same MFP performance is caused by much higher signal-to-surface-generated noise ratio (SSNR) than signal-to-interference ratio (SIR). 3) MSE at high SNR deceases in a slower pace with surface-generated noise than the case with white noise or a point interference.