Recognition of ship echo signal applying wigner distribution feature

Abstract

Recognition of ship echo signal's difficulty lies in extracting effective feature from target reflection signal. Researches show that target echo signal is double grading with time and frequency, so time-frequency is a effective way in the field of echo signal recognition. Wigner distribution is a real means of time-frequency analysis, but it is limited by its false time-frequency spectrum called interference item. In the paper, on the basis of studying the theory of interference item generation and reduction, Choi-Williams kernel function and self frequency-window methods are applied to implement Wigner's interference reduction. Through simulation test using typical signal such as multicomponent signal, Merits and drawbacks of the two means are also analyzed. Then to merchant ship, reef and reverberation, three kinds of CW echo signal, the interesting time-frequency features are extracted. At last, sample sets of above three kinds of echo signals are divided into training and testing sample sets. The number of training samples to the number of testing samples ratio is 1 to 4. The training samples are regarded as typical sample and input to Fuzzy Adaption Resonance Theory (FART) network to train. According to typical samples, the testing samples are tested by the same network. The results show that self frequency-window is a better means to reduce interference item. But high recognition rate can be achieved if interference items exist.