基于字典学习的主动声呐目标分类方法

主动声呐目标分类在军事和民用方面都有重要的应用和价值。文章基于稀疏表示理论，结合K-奇异值分解和正交匹配追踪算法，提出一种基于学习字典的稀疏表示分类方法（Dictionary Learning Sparse Representation Classification，DLSRC）。首先，利用K-奇异值分解算法训练各个类别目标回波信号，得到带有目标特征信息的类别字典，类别字典对信号具有良好表征能力并且带有目标类别信息；然后，利用正交匹配追踪算法和各个类别字典稀疏分解测试信号，得到各个类别字典下的稀疏系数后重构信号；最后，根据各个重构信号与测试信号的匹配度判定类别，得到分类准确率。结果显示，200个测试数据在信噪比分别为-5、-3、6 dB时，DLSRC法的分类准确率分别达到87%、89%、95.5%。不同信噪比下基于学习字典稀疏表示分类方法的准确率均高于已有的支持向量机（Support Vector Machine，SVM）、K-最近邻（K-Nearest Neighbor，KNN）和柔性最大值分类器（SoftMax）等分类方法，具有较好的分类性能。

Active sonar target classification based on dictionary learning

Active sonar target classification has important applications and values in both military and civil fields. By using sparse representation theory and combining K-singular value decomposition algorithm with orthogonal matching pursuit algorithm, a sparse representation classification method based on dictionary learning sparse representation classification (DLSRC) is proposed in this paper. Firstly, K-singular value decomposition algorithm is used to train the echo signals of each category of targets and to obtain the category dictionary with target characteristic information, which has good representation ability to signal and contains target category information. Then, the orthogonal matching pursuit algorithm and each category dictionary are used for decomposing the test signals sparsely to obtain sparse coefficients under each category dictionary and to reconstruct signals. Finally, according to the matching degree of each reconstructed signal and reconstruct signals. Finally, according to the matching degree of each reconstructed signal and test signal, the classification accuracy is determined. The results show that when the SNR of 200 test data is -5, -3, and 6 dB, the classification accuracy of DLSRC method for the 200 test data reaches 87%, 89%, and 95.5% correspondingly, which is higher than that of the existing support vector machine (SVM), K-nearest neighbor (KNN) and SoftMax classification methods.