Recognition of Acoustic Emitted from Surface Vessels Using MobileNet Convolutional Algorithm

With the movement of the vessels in the water and the activity of the propulsion engines and the rotation of its propellers, they emit sound signals from them, which are called the ship's radiated noises. Today, the naval forces of the world use these sounds to identify surface vessels passing through territorial and international waters. One of the best methods for classifying and recognizing vessels according to the sounds emitted by them is deep learning. By using deep learning, it can extract the unique features of the signal, which have high accuracy in recognition. This paper designed a model based on the Mobilenet network, which processes the acoustic signals received by underwater sound receivers (hydrophones) and finally classifies them with high accuracy. The input of this model is the spectrogram images related to passive sonar sound data, which are produced using short-term frequency transformation (STFT). We created this model in the Python program using the keras library and the results show that the accuracy of the proposed model is more than 96% and its evaluation loss is less than 3%. Compared to the common methods of deep learning, the proposed method, in addition to having a suitable calculation speed, also has an acceptable recognition accuracy.