On The Robust Detection Methods And Its Applications On Underwater Signal Processing

Robust detection theory can be come back to the paper published by fluber in 1965, it has been developed quickly and has many practical applications in various fields for its robustness against deviation from assumed models and against error as well as other robust signal processing methods. For inevitable interference and error under complex ambient backgrounds, conventional detectors for weak signals will not work well. With the development of signal processing techniques and fast algorithms, the signal detection theory got much progress. Generally speaking, robust detection methods can be divided into two major branches, classical methods which are based on robust hypothesis testing and modern methods based on new signal processing methods.This paper first reviews the evelopment history of robust detection theory and its applications in underwater acoustic signal processing, and then following the introduction of main features of ship-radiated noise and ambient noise. Based on the theory and knowledge, the paper proposes and discusses several robust detectors in detail, such as optimal periodical signal detectors which exploits the prior knowledge that the chi_square distribution hypothesis test has certain robustness, detectors based on over_sampled Gabor representation for transient signals detection which takes the numerical computational advantage of robustness brought by over_sampling, and the detector using computer_intensive method "Bootstrap". Furthermore, based on the idea from Hinich, presents a novel concept of "signal coherent function" which measures the stability of signal components. The results from computer simulation shows that the proposed detectors are practical and effective and have wide application prospect. Finally the author points out the difficulties existing in practical applications, research directions and the development trend of robust signal detection in the future.