Enhanced Event Detection with Seismic Listening Sensors

Physical Sciences Inc. (PSI) proposes a Phase II program to develop, test and implement in a hardware/software system an advanced set of statistics-based algorithms for detecting anomalous events for a wide range of seismic activity levels. Phase I quantitatively demonstrated the feasibility of our statistics-based approach, yielding event detection with p_d > 90%, P_fa 20. Phase II will develop algorithms that can analyze a long period of signal from an environment, measure its activity level using entropy and multifractal dimensions, analyze its trends, learn the relationships between states of the signal through statistical learning methods, and then detect the time and location of unusual events using innovative statistical decision making techniques. These algorithms will be capable of detecting low level (e.g., a rural area) and high level (e.g., marketplace) seismic activities in the background and adjust their thresholds, metrics, and decision making methods for increasing their probability of detection to above 95%, reducing their false alarm to less than 1%, and location detection error to less than 30m. In addition, these algorithms will reduce the required storage space for the signals by at least 2 orders of magnitude.