Design And Implementation Of Handwritten Digital Recognition System

In recent years, handwritten digital recognition takes advantage of the computers to identify handwritten digital technology and it is widely used in the current character recognition research. The efficiency and reliability are the keys to the design of system. According to the general character recognition technology, our system can be divided into three important parts: the processing of handwriting digital image; the extraction of feature; the training and recognition of BP neural network.The processing of handwriting digital image can be divided into preprocessing and character segmentation. The preprocessing takes use of common used method combined with the particularity of the character to design an efficient process. After preprocessing, the image can remove noise and interference so as to prepare for feature extraction. Character segmentation technique is applied to mobile phone number identification processing. We need to use the vertical projection segmentation to save these 11 images as an independent image, and then normalize these digital images to the standard image space.Feature extraction is another important problem in the process of digital recognition. Through comparing with various identification methods and combining with the image characteristics, we take advantage of the coarse grid extraction method to extract features which proves this method the ability to deal with noise of image and it is applied to the features of fixed stroke. Coarse grid extraction method improves the recognition rate which proves the extraction effect is good.Neural network classification equipment has the ability of learning and training, and it has realized the classifications of sample images which makes the computer simulation of the process of brain. BP neural network has been used to handwritten digital recognition technology and the result proves the feasibility of system.Based on the application of the above processing step, we simulate the entire identification process by code that comprehensively reflects the characteristics of handwritten digit recognition. Through experimental test, the system is feasible from the process of pretreatment to the classifier design and achieve good recognition rate.