COGNITIVE PASSIVE SONAR CLASSIFICATION USING INTEGRATED MULTI-CONVOLUTIONAL NEURAL NETWORKS AND DATA FUSION

Abstract

Currently, machine learning and artificial intelligence algorithms are used to perform multiple tasks to improve efficiency and accuracy of outputs. In the area of passive sonar target classification, many machine learning techniques have been studied to improve the quality target of recognition. The goal of this thesis was to evaluate whether improved feature extraction and target classification can be achieved by combining multiple convolutional neural networks with data fusion. We produced a cognitive passive sonar algorithm and assessed its effectiveness and applicability in acoustic signal processing. Our research has demonstrated that the integration of three convolutional neural networks combined with data fusion can substantially enhance the functionality of existing passive sonar systems. This architecture was able to produce reliable outputs that improved the accuracy of decision-making. Furthermore, the algorithm has the potential to scale across other machine learning techniques, making it robust enough to be expanded for future improvements.