Classification and Identification of Underwater Small Target Based on Multi-Parameter Joint Feature

The active sonar echoes contain a large amount of target information. The target can be classified and identified by feature extraction. Aiming at the problem that underwater small target cannot be classified very well based on a single feature, a classification and identification method of underwater small target based on multi-parameter joint feature is proposed in this paper. Different forms of transmit signals such as linear frequency modulation(LFM) signal, hyperbolic frequency modulation(HFM) signal and bat biomimetic signal are selected to combine the variance, the spectral centroid and the peak of wavelet energy spectrum of the target echo, and the back propagation(BP) neural network classifier are used to classify and identify the targets. Pool experiment shows that the classification and identification method of underwater small target based on multi-parameter joint feature can effectively improve the target classification identification rate, and compared with traditional signals, using the bat biomimetic signal as transmit signal can obtain a higher classification identification rate. This work may provide a reference for making use of multiple features to jointly classify and identify underwater targets.