Passive Acoustics Sonobuoys Intelligence and Machine Learning

SSC proposes to develop Concurrent Autonomous Multi-stream Passive Fusion for Intelligent Reconnaissance Engagements (CAMPFIRE), an innovative combination of machine learning and Bayesian fusion models for passive acoustic processing that enables high levels of automation for Directional Frequency Analysis and Recording (DIFAR) sonobuoy fields. Our approach is based on state-of-the-art transformer deep learning models that detect and classify submarine energy signatures in bearing-colorized spectrograms. These models operate concurrently on many signal processing configurations (resolution codes) to maximize the Signal-to-Noise Ratio (SNR) of individual components. Signature detections from different signal processing streams are combined at the buoy level and passed to a Bayesian fusion engine that associates energy components across the DIFAR field and tracks contacts based on their bearing and classification scores. Tracks consistent with submarine activity will generate automated purple reports for highly probable threats and a field level scissor gram that combines the components