Computationally Efficient Linear Transforms for Next-Generation Remote Sensing Systems

University Of Oklahoma Norman Campus, Norman OK

Investigators

Abstract

The objective of this research is to develop a new technique for finite bit representation to enable high-performance signal processing for remote sensing systems. The technique for finite bit representation is based on multidimensional continued fractions coupled with an architectural design strategy that does not use multipliers. Multidimensional continued fractions have been known within the mathematical community for some time, but have largely been unexplored by the engineering community until recently. Simultaneous rational representations are a member of the multidimensional fraction family and are employed to create fixed integer transforms with computationally optimal representations. This research seeks to provide an approach that can decrease the size, weight, and cost of sensor systems and increase their spatial and temporal resolution. Various hardware platforms are to be explored during the project. Of these, field programmable gate arrays and digital signal processors are likely the most compelling for fast, low-power devices. The principal investigator is currently serving as the faculty advisor for the American Indians in Science and Engineering Society chapter at the University of Oklahoma and the co-principal investigator is currently serving as a mentor for two high-school robotics teams near Oklahoma City. In addition, the students from the Society of Women Engineers, and the 436 students in the university?s Multicultural Engineering Program are encouraged to participate in this project. Remote sensing systems include satellite, light detection and ranging (LIDAR), infra-red, and radar. As such, this project helps students and researchers learn more about the environment in which they live. Experiments utilize the National Weather Radar Testbed that is located in Oklahoma.