Machine learning-based classification of impulsive acoustic sources in urban settings using ground-coupled airwaves

This study examines the feasibility of using ground-coupled airwaves for detecting and classifying impulsive acoustic sources in complex urban environments. Ground-coupled airwaves, generated by high-energy acoustic events such as gunshots and explosions, interact with the ground surface and propagate as seismic waves. A distributed seismic sensor array is employed to capture these signals, which are shown to travel at velocities near the speed of sound in air and dominate the recorded waveforms. Time–frequency analysis combined with deep learning techniques enables effective classification of source types based on their distinct signatures. Results demonstrate that ground-coupled airwaves retain sufficient discriminative features for reliable classification, even under the reflective and scattering conditions typical of urban terrain. By integrating seismic sensing with modern machine learning, this work contributes to developing reliable, scalable systems for threat detection and situational awareness in urban surveillance, environmental monitoring, and security applications.