DSP Implementation Of Offline Digits Recognition Technology Based On Neural Network

Character is an important communication tool in our life. It is simple when looked as a figure. Character figure is easier to be recognized, processed and classified than other figures, such as face figure. Much complex work, such as reading data, classifying post, recognizing all kinds of report forms, will be finished automatically if character recognition technology is used. The main objects of character recognition are letter, digit, symbol and Chinese character. There are two recognition methods: online recognition and offline recognition. Design and implementation of offline digits recognition is the main task of this project.The task of this project can be divided to three parts: image sampling, image pre-processing, image feature extracting and recognizing. Several devices including image sensor, CPLD, video decoder, FIFO, are needed to finish image sampling. Video signal is decomposed to control signal and digital image signal by video decoder. Digital image data form video decoder will be saved in FIFO under the control of CPLD.Image pre-processing including image smoothing, target enhancement, image thresholding, make image easier for feature extracting and recognizing. A target enhancement algorithm is put forward in this project. It is proved that it is much easier to extract image feature by employing target enhancement algorithm.Image features to be extracted depend on the need of image recognition. Neural Network can finish adaptation, organization and learning automatically. It is used abroad in the field of image recognition. Neural Network was used to recognize digits in this project. Display module and communication between DSP and PC module are also designed in this project.Hardware design and debugging of offline digits recognition system were finished. Image processing and recognition algorithm was also implemented in DSP environment. In this paper, the basis theory of this project was introduced firstly. Then , the details about design and debugging were introduced. Finally, some measures are discussed to improve system efficiency. This system will be of more application value after further mend.