The System for Handwritten Symbol and Signature Recognition Using FPGA Computing

The system for handwritten symbol and signature recognition with the use of shape descriptors in orthogonal basis of Hadamard transform as informative features is presented. For a construction of the classifier discriminant methods have been used that have supplied acceptable separation between classes. To raise processing speed a part of algorithmic scheme was implemented by hardware facilities. There was used specialized integer-valued co-processor implemented as PCI bus extension board based on FPGA. Proposed hardware/software architecture resulted in computational acceleration factor of 10 with recognition rates averaged about 97-99% in the task of handwritten symbol processing.