Integrated computing system for real-time data processing, storage and communication with large aperture 160-element coherent hydrophone array

Underwater acoustic data contain signals from a myriad of sound sources and monitoring these acoustic events is important for many ocean applications. Substantial volumes of underwater acoustic data are usually acquired in passive ocean acoustic waveguide remote sensing (POAWRS) experiments with a large-aperture densely-populated coherent hydrophone array system. Developing automatic, accurate and fast algorithms for the detection and classification of underwater acoustic events can help to minimize the human effort, enabling real-time processing and analysis, and hence aiding rapid scientific discoveries at sea. We utilize acoustic data from a 160-element coherent hydrophone array and beamform this data in 147 distinct bearings. To speed up the data processing, we investigate a variety of accelerating approaches, such as analyzing the effect of floating point precision, applying parallel processing, and implementing fast algorithms to run on GPU. We implemented and optimized delay and sum beamforming, both in the time domain and Fourier transform domain on NVIDIA GPUs, and achieved more than 338x speed-up compared to the basic CPU-implemented version. All of the coherent hydrophone array data preprocessing stages, including beamforming and Direction of Arrival (DoA) estimation are calculated in real-time. During sea trial of the Northeastern University (NU) inhouse designed and fabricated 160-element coherent hydrophone array hardware at the U.S. Northeast coast on board the research vessel (RV) Endeavor in September 2021, we utilized the developed and optimized software, as well as high-performance computing hardware, specifically configured and assembled inhouse, to store, process and analyze data sampled at 100 kHz per hydrophone element in order to monitor acoustic events in real time. Because of the high sampling rate, we could monitor a large variety of sound sources, from high frequency events such as dolphin echolocation clicks with significant energy near 50 kHz, down to low frequency fin whale 20 Hz pulses.