Development of synthetic infrasound signature models from window of opportunity observations

Between May 21 and May 22, 2024, researchers from Georgia Tech Research Institute deployed an infrasound sensing array at Cabrillo National Monument and by the shoreline at Torrey Pines, San Diego, California to measure the infrasound signatures of sources of interest such as rotorcrafts, jets, and large boats. This source library was developed to facilitate aircraft detection and classification using infrasound measurements. Pictures of potential targets with the corresponding time-stamps were taken for visual identification and data annotation. In post-processing, the recorded time-series from the infrasound array were associated with the target pictures and delay-and-sum beamforming was used to maximize the signature signal-to-noise ratio (SNR) across the observation window. Using the spectral characteristicsfrom the annotated data, a source signature model capable of generating the time-series for an acoustic source in motion relative to a receiver was developed and tested. Synthetic acoustic data constructed using the source signature model were used to augment the sparse measurement dataset, with the goal of improving target classification performance. The effectiveness of this approach was assessed by comparing the results of classifiers trained with and without dataset augmentation to identify the class of the source signature, i.e., the type of vehicle producing that signature.