An Identification Method of Underwater Targets Based on Sparse Representation

An identification method of underwater targets based on sparse representation model with mixed-norm regularization is proposed in this paper. The proposed model employs three different features of acoustic signals, which are with complementarity and correlation: the central moments feature, the wavelet packet component energy (WPCE) feature and the Mel Frequency Cepstral Coefficients (MFCC) feature. From those features of training data, a sparse representation matrix can be optimally estimated. Then, the class labels for test samples are determined via the minimum reconstruction error criteria. To evaluate our model, a pool experiment of three different targets has been conducted, and the results show that the proposed method has high recognition accuracy.