Abstract
Underwater acoustic target recognition based on ship-radiated noise belongs to the small-sample-size recognition problems. A competitive deep-belief network is proposed to learn features with more discriminative information from labeled and unlabeled samples. The proposed model consists of four stages: (1) A standard restricted Boltzmann machine is pretrained using a large number of unlabeled data to initialize its parameters; (2) the hidden units are grouped according to categories, which provides an initial clustering model for competitive learning; (3) competitive training and back-propagation algorithms are used to update the parameters to accomplish the task of clustering; (4) by applying layer-wise training and supervised fine-tuning, a deep neural network is built to obtain features. Experimental results show that the proposed method can achieve classification accuracy of 90.89%, which is 8.95% higher than the accuracy obtained by the compared methods. In addition, the highest accuracy of our method is obtained with fewer features than other methods.
Highlights
A passive sonar system is the main equipment for identifying underwater or surface targets through their radiated noise
By applying the feature-selection algorithm on traditional features, the support vector machine (SVM) classifier could achieve an accuracy of 83.42% with 36 features, while the highest accuracy obtained on competitive deep-belief networks (CDBNs) features is 90.89% with nine features
It is found that the deep-belief network pretrained with a large amount of unlabeled ship-radiated noise can solve the problem of the lack of training samples
Summary
A passive sonar system is the main equipment for identifying underwater or surface targets through their radiated noise. We propose a new DBN method called competitive deep-belief networks (CDBNs) for underwater acoustic target recognition. The main idea of the proposed method is: (1) pretraining the DBN with a large amount of unlabeled data in an unsupervised manner to learn basic concepts of underwater acoustic signals; (2) grouping deep features according to their relevance with categories;. The proposed CDBN method integrates a new competitive learning mechanism into deep-belief networks to learn more robust and discriminative features for underwater acoustic target recognition. The experimental results demonstrated that the proposed CDBN method is effective for underwater acoustic target recognition It can significantly reduce the random noise and enhance the line-spectrum characteristics of ship noises, and the CDBN features have better classification performance than other hand-engineered features.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have