ANDRO COMPUTATIONAL SOLUTIONS LLC — Department of Defense SBIR Phase I: A19-129

Description

Tactical signal intelligence (SIGINT) systems are often presented with an amalgam of radio frequency (RF) signals emanating from friendly, hostile and neutral sources. Elucidating the electromagnetic spectrum (EMS) by identifying the resident signals will provide the operators with actionable intelligence. Ability to identify the myriad RF waveforms that the SIGINT will plausibly face is key to differentiate and identify threats in the EMS. To this end, ANDRO Computational Solutions, LLC will design, develop and prototype first of its kind small form-fit, multi-task Machine Learning (ML) based SIGINT device – DeepSPEC: Artificial intelligence (AI)-powered blind signal detector and classifier – that employs FPGA for ML acceleration. Unlike the energy inefficient GPU platforms, FPGA based accelerators are gaining momentum for low size, weight, and power (SWaP) platforms due to their higher power efficiency (i.e., performance per Watt), parallelism for real-time performance, reconfigurability, and low development cost. DeepSPEC will possess AI-powered signal detection, characterization, advanced RF visualization, and cataloging for a comprehensive and diverse RF database on a low SWaP COTS platform.