Passive Sonar Detection and Classification Based on Demon-Lofar Analysis and Neural Network Algorithms

This paper focuses on an experimental study that used passive sonar sensors as the primary information source for the submerged target in order to identify, classify, and recognize naval targets. Surface vessels and submarine generate a specific sound either by propulsion systems, auxiliary equipment or blades of their propellers, producing information known as the "acoustic signature" that is unique to each type of target. Consequently, the analysis and classification of targets depend on the processing of the frequencies produced by these vibrations (sound). utilizing the TPWS (Two-Pass-Split Windows) filter, this work aims to develop a novel technique for target identification and classification utilizing passive sonars. This technique involves processing the target's signal in the time-frequency domain. subsequently, in order to improve the frequency lines of the target noise and decrease the background noise, a TPSW algorithm is implemented in the frequency domain. By integrating narrowband and broadband analysis as inputs of an artificial intelligence model that can classify a target into one of the categories given in the training phase, the target has finally been classified. Our findings demonstrated that the suggested approach is dependent upon the size of the target noise data collection and the noise-to-effective-signal ratio.