A New Depth Classification Method based on Underwater Acoustics for Naval Defense Applications

Abstract

The main purpose of this research is to present an automatic underwater acoustic classification model with high performance. Thus, a new sound dataset was collected. By using this dataset, a new underwater depth classification method is proposed in this work. Average pooling has been used to pre-processing underwater sounds. The used average pooling model is both removed the noises and compressed signal. S-transform and AlexNet have been used for feature extraction. By deploying S-transform to underwater sounds, contour images have been obtained. These images have been utilized input of the AlexNet. Herein, AlexNet has been utilized to extract features by using transfer learning. Features extracted have been classified with the Support Vector Machine (SVM). In our method, 99.05% accuracy has been calculated. The calculated results and findings obviously illustrate the success of our proposed S-transform and AlexNet based model on the underwater sound classification.