SONALYSTS INC — Department of Homeland Security SBIR Phase I: DHS221-001

Description

The US Coast Guard (USCG) Search and Rescue (SAR) program’s mission is to prevent loss of life in every situation where actions or resources could possibly be brought to bear. Because of this serious obligation, the USCG expends resources to pursue all distress notifications until they can be identified as a hoax. Sector Communications Unit (CU) watchstanders typically spend 12- hour watches listening mostly to radio static, which can lead to calls being missed. Further, when a distress call is received, it can be difficult to determine its authenticity. In this Phase I effort Sonalysts, Inc. will conduct research and development of the Heuristic-Optimized Analysis of Xmissions (HOAX) system, a human-centered Artificial Intelligence (AI) that will augment watchstander capabilities rather than seek to replace them. We will fuse various streams of data in addition to the voice transmission itself to provide watchstanders with classifications of critical cues (e.g. slurred speech or laughter), rather than a binary “actual” or “hoax” classification that must be accepted or rejected in totality. AI models will aid OPCEN watchstander sensemaking by helping them understand if certain cues are present or absent, while providing various overlays of additional geospatial data. To achieve this vision we will start by conducting user research with CU watchstanders, SAR Mission Coordinators (SMC), Sector OPCEN Chiefs, and Sector Commanders to identify critical cues used in hoax determination. We will conduct formative analytics to see which of these cues can be reliably classified to shape a research plan for Phase II.