This project will develop methods to efficiently identify, localize, and track in space and time a threat event using “data streams of opportunity.” Of specific interest is sensor data from mobile devices including accelerometer, audio, and perhaps video. We envision a scenario in which members of a community (e.g., school, commuters at a transit station, etc.) opt into the use of a cell phone app capable of collecting and transmitting information to a Virtual Sentry. Just as Google employs data from their Maps users to predict traffic patterns and suggest routes, we will use the data provided by the community to detect and characterize threats. As the project progresses, we will consider adding to the models and algorithms information from other venue sensors such as spectral/hyperspectral signatures, chemical “sniffers,” or millimeter/THz anomaly detectors.
Our approach will build on and extend methods in probabilistic modeling and processing, reduced order signal representation, and dynamic sensing. We will make no assumptions concerning the number or types of sensors, the manner in which these sensors are distributed across space, or how the data are sampled in time. The framework will be broadly applicable across a range of threat scenarios and will support a wide range of fusion architectures. This project will provide DHS with a principled, adaptive approach to the acquisition and processing of heterogeneous streams of data for threat identification and dynamic characterization.
