Underwater acoustic targets recognition algorithm based on NMF

Abstract

To solve the problem of low recognition rate of underwater targets for the reason of their large feature discreteness within class and high feature overlapping between classes, the underwater targets recognition algorithm based on NMF universal dictionary model (UDM) is proposed, in which, the UDM is established using the existing underwater acoustic target data, and the identification matrix is also established at the same time using the weight coefficient matrix of each kind of target, and then, the block sparse nonnegative matrix factorization is carried out to the unknown underwater targets and the classification and recognition of underwater acoustic targets are achieved by comparing the weight coefficient matrix and the identification matrix, compared with another three methods based on measured data experiments, the proposed algorithm achieves better classification results.