A parallel relaxation architecture for handwritten character recognition

Though probabilistic relaxation is a class of iterative, parallel techniques for image processing, pattern recognition and so on, it sometimes takes an unacceptable time to solve a given problem on a conventional sequential machine. In the paper, a parallel architecture HANS which consists of a host computer and multiple relaxation processors (RP) is proposed. HANS accelerates not only the relaxation operations, but also preprocessing for relaxation operations to accelerate all processes and reduce the communication required between the host computer and the RPs. The RP has DSP for main processing, memories and dedicated hardware specialized to the recognition of handwritten Japanese characters. The performance evaluation of the proposed architecture is also shown