The DSP Implementation Of Handwritten Numeral Recognition Algorithm

Pattern recognition is a hot area of international study. The merit of a pattern recognition algorithm can be judged by the level of IRIS data detection and hand-written numeral recognition. Hand-written numeral recognition is considered as same difficult as Chinese character recognition. It is not only used to judge the merit of a recognition method, but also that the study on it greatly promotes the research of recognition method, and extends the pattern recognition field. So the hand-written numeral recognition is an important study subject in pattern recognition field. High-speed, high-precision numeral recognition system can replace the manual data entry, the identification of the postcode and bank notes, and other heavy work, which greatly improve the speed and efficiency of work. It has broad application prospects.In the image analysis, FCM clustering algorithm can be effectively used for image processing, especially for hand-written numeral recognition. However, because of the complexity of hand-written numeral recognition, how to design a better solution, and how to achieve recognition better through the software, become a research focus.The main task of this paper is to design an off-line handwritten numeral recognition device which is high-precision, high-speed, small size and low cost. According to the requirements of handwritten numeral recognition algorithm, the hardware for handwritten numeral recognition system is designed, which includes: image acquisition module, image processing module, memory module and power supply module. Image acquisition module uses the Philips SAA7115 as image acquisition chip, and image processing module is composed mainly by the TI's TMS320VC5416. Image acquisition chip digitized analog signal which was collected, then sent it to RAM through FPGA. When an image was deposited over in the RAM, it would be sent to DSP for real-time processing, and then the result would be delivered to PC for treatment. Throughout the system, the FPGA is used to realize the logic connections, communication and response between the major chips, initialize the image acquisition chip, and generate the memory address where the data were sent to in RAM. Memory module is used to store user programs and image data.The handwritten numeral recognition algorithm used in this paper is briefly introduced and discussed, according to algorithm requirements, the hardware system is designed, and the schematics of circuits are preliminary completed. The system's software design diagram is given, as well as the flowchart of I2C data transfer program. Finally, a general summary of the entire system is given, and the problems encountered at the design stage are pointed out, as well as the solutions adopted; the places which can be improved in the system are pointed out, and the future direction of work is set out.