Classification of anti-submarine warfare sonar targets using a deep neural network

A well-known problem with modern anti-submarine warfare sonars with narrow beamwidths and wide frequency bandwidths, is the frequent occurence of false alarms, particularly in littoral environments. This increases the workload of sonar operators and also reduces the usefulness of automatic systems such as autonomous underwater vehicles, since their limited communication abilities hinder them from sharing large amounts of contacts. In this paper, deep learning is applied on sonar data with a high amount of false alarms. Even though the data set is quite small, it is shown that it is possible to train a deep neural network that outperform simple signal-to-noise ratio (SNR) thresholding. Due to the limited amount of available data, a simple augmentation scheme is employed to increase the training and validation data sets, by copying some of the training instances. This augmentation scheme is shown to significantly improve the results of the training.