Robust MUSIC-like array processing for underwater applications

The content of this thesis contains original research on MUSIC-like array processing techniques for underwater sonar applications. The main objective of this research is to develop robust algorithms for super-resolution performance in sonar systems and has contributed novel algorithms that solve several problems that super-resolution processing encounters in underwater sonar applications. Results of detailed analysis and validation of algorithm by real data are also reported in this thesis. The Generalized MUSIC-like algorithm is proposed to cope with the demanding and complex noise situations encountered by underwater sonar systems by using a quiescent covariance matrix model. The solution is shown to be able to mitigate the effects of complex underwater noise in simulations and real data processing. The Doubly Constrained Generalized MUSIC-like algorithm is proposed to mitigate the adverse effects of steering vector error by using error model in double quadratic constraints optimization problem. A novel method is introduced in this thesis to solve the double-quadratic-constraints problem, which does not perform linearization. Simulations and data processing results show that the algorithm is effective with remarkable improvements over the Generalized MUSIC-like algorithm. The Generalized MUSIC-like algorithm is extended to the acoustic vector sensor array configuration to achieve high performance resolution and left/right ambiguity rejection amid much more complex quiescent condition owing to its non-uniform sensing mechanism. Real data processing results show excellent resolution and ambiguity rejection performance from the proposed algorithm with both stationary array and towed array. The Diagonal Loading Generalized MUSIC-like algorithm is proposed to mitigate the effects of non-positive definite quiescent and covariance matrices; and mismatched quiescent covariance matrix model for acoustic vector sensor array. Results from simulations and real data processing show that the proposed algorithm is capable of mitigating the effects of mismatched quiescent covariance matrix model and obtain substantial improvement in resolution and ambiguity rejection performance. The Generalized MUSIC-like algorithm is further extended to the hybrid array configuration. Real data processing results show that the proposed approach and configuration yields better performance over full acoustic vector sensor array which faces more challenges in extremely challenging environment and maintaining uniform sensors orientation. The Generalized MUSIC-Capon algorithm is proposed for active sonar applications to overcome the difficult over estimating quiescent covariance matrix in both spatial and frequency domain. The algorithm is used to process real experiment data and is shown to be effective in mitigating the effects of reverberation. The range-bearing and range-Doppler spectra show that the Generalized MUSIC-Capon algorithm performs much better than both the conventional and Capon’s method in terms of target detection and Doppler estimation with its super-resolution performances.Doctor of Philosophy (EEE