INTEGRATED SENSING: Distributed Multiscale Geometric Processing On Real-Time Sensor Networks On Radiation Fields

Duke University, Durham NC

Investigators

Abstract

0225414 Brady A group of engineers specializing in multiplex sensors and computer scientists specializing in computational geometry will jointly develop physical and logical geometric processing to improve sensor network efficiency. The efficiency of sensor system may be defined as a ratio of the number of sensor states to the number of system output states. An efficiency ratio near 1 dramatically reduces the computation and communication costs of sensing. Reduced computational loads are critical to real-time operation. Multiplex sensing (meaning non-isomorphic mapping between the source state and sensor pixels) is the critically enabling feature of efficient systems. Geometric structure is used to encode multiplex sensitivity. This project may lead to motion and tracking sensors for human-computer interfaces at much lower cost and with much lower computational complexity than current technology. This project will leverage sensor network testbeds developed through the Duke Information Spaces Project (DISP), including Argus, a 128 video camera array on a 14 meter ring connected in parallel to a Beowulf processing cluster, 5,000 square foot pyroelectric and acoustic sensor volume in the Smith tobacco Warehouse and ad hoc sensor networks in the Duke Forest.