SBIR/STTR Award attributes
Vadum will develop MATCH (Multimodal AcousTic Characterization of Hoax calls) a novel machine learning (ML) system to automatically and accurately detect deception in maritime Search and Rescue (SAR) audio distress calls and aid Coast Guard watchstanders in disambiguating hoaxes from legitimate calls. MATCH will substantially alleviate the burden on Coast Guard watchstanders by continuously monitoring the VHF distress channel for traffic, analyzing detected distress calls, and issuing warnings and alerts when hoax calls are present.MATCH will incorporate expert-identified and learned features that profile the acoustic signature of the caller, lexical traits, and characteristics of the acoustic scene (background environment) to confirm or invalidate the authenticity of the distress call.Legitimate distress calls typically happen in a specific acoustic scene.MATCH will utilize a novel classifier to recognize and validate legitimate acoustic scenes. The multi-mode multi-feature nature of the MATCH architecture will accommodate variances in regional dialects and accents and be expandable to include additional features representative of intoxicated mariners. The MATCH solution is adaptable to other domains such as 911 call centers, law enforcement agencies and commercial voice phishing attack mitigation.
