Smart Sensing and Recognition Based on Models of Neural Networks
Abstract
Radar target identification is an important problem to which much attention has been given in the past solution efforts relied predominately on linear signal processing techniques. There are two traditional approaches. In the one, high resolution images are formed to be examined and identified by human observers. In the second, target signatures (feature vectors) are formed for automated machine identification. The first approach is usually quite costly and has several practical limitations stemming from the high cost and large size of microwave imaging apertures. The second approach is yet to provide a reliable scheme. Motivated by the observation that the above approaches are primarily linear and that biological systems, which process information in a highly nonlinear, collective, and frequently iterative manner, are very adept at carrying out recognition, classification association, and optimization tasks, we elected to investigate the capabilities co collective nonlinear processing in target identification.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 15, 1990
- Accession Number
- ADA230701
Entities
People
- N. H. Farhat
Organizations
- University of Pennsylvania