Electromagnetic Signal Classification Using Ensembles (E-CUE) Navy SBIR FY2010.3 Sol No.: Navy SBIR FY2010.3 Topic No.: N103-224 Topic Title: Electromagnetic Signal Classification Using Ensembles (E-CUE) Proposal No.: N103-224-0150 Firm: Charles River Analytics Inc. 625 Mount Auburn Street Cambridge, Massachusetts 02138-4555 Contact: Wayne Thornton Phone: (617) 491-3474 Web Site: www.cra.com Abstract: US military forces have encountered increasingly complex electromagnetic (EM) environments over the last two decades. This complicates the task of rapidly and properly classifying emitters, which is critical to the safe, effective operation of combat platforms. Many large military platforms such as warships and aircraft rely on electronic warfare support (ES) systems for self-defense. Platforms conducting missions such as covert strike, cooperative targeting, and tactical engagement (e.g., anti-surface warfare (ASUW)) require ES capabilities in addition to those required for self-defense, since they must intercept, identify, and locate or localize emitters of interest. To rapidly and correctly fingerprint sophisticated emitters, we propose to design and demonstrate an ES system classification tool, Electromagnetic Signal Classification Using Ensembles (E-CUE). E-CUE incorporates several novel capabilities, including signal features based on alternative characterizations, classification ensembles, characterization of classification uncertainty and reliability, emitter correlation and fusion, adaptability through process assessment (PA) and process management (PM), and improved early warning. Benefits: E-CUE technologies will be initially targeted at military applications for sensor processing-for electronic warfare (EW) as well as for other types of sensor processing systems. E-CUE technologies will also be useful in any domain or application requiring adaptive classification of high-dimensional data. Examples of these applications include sensor processing (e.g., computer vision, image processing, and video surveillance); speech recognition; text recognition (hand-written, on-line, and printed); and even laboratory analysis of DNA microarray data. In addition, we plan to transition specific classification and fusion algorithms and features of the overall system architecture to our VisionKitT software suite to increase its commercial viability. Return