Underwater Target Classification Using Deep Neural Network

Underwater target classification is one of the most popular topics in underwater signal processing. A lot of different methods have been proposed to deal with this problem. However, most of them rely on human experience to extract low-level features. In this paper, a novel classification system based on deep neural networks (DNN) is proposed, which utilizes the strong modelling ability of DNN to learn high-level features from wavelet packet component energy (WPCE) features automatically. We aim to use DNN to generate more discriminative high-level features with the WPCE, since the wave packet transform can offer a richer range of detail characteristic than spectrum analysis. Our experiments on the real recorded data of five type of vessel indicate that the proposed system using WPCE features provides higher classification accuracy than the system using spectrum features.