課題別成果報告 -詳細-

Developing a Cognitive System for Pattern Recognition and Hardware Implementation in FPGAAli Ahmadi, Tetsushi Koide, M. Anwarui Abedin and Hans Jurgen Mattausch

Abstract- During my research period under the COE program, I have been working on two projects in parallel. First, is developing an associative memory based learning model which uses a short and long-term memory similarly to the memorizing procedure in the human brain. The model uses a ranking mechanism to control the reference patterns life-time and an optimization algorithm to adjust the reference vectors components as well as their distribution, continuously. Comparing to other learning models like neural networks, it is advantageous in term of no need to pre-training phase as well as its hardware-friendly structure which makes it implementable by an efficient LSI architecture without requiring a large amount of resources. The system is already implemented on an FPGA platform and tested with real data of handwritten and printed English characters delivering satisfactory results. The second project is a hardware implementation for Active contour models (Snakes) which is used for moving object detection and video tracking. I have proposed a new algorithm for snakes hardware implementation which is based on use of a number of parallel cell units for snake points and a main controller for all control tasks and main memory access. Using this parallel architecture we could obtain a very reasonable snake generation time (16ms) for VGA image size which was used for motion detection in video samples of 24 fps in a simulation program.