| Handwritten digit recognition has wide application prospect in many fields, domestic and foreign scholars doing a lot of research about this, and brought up many pretreatment and pattern recognition algorithm, which greatly improved the accuracy of the handwritten digit recognition. But so far, the accuracy of handwritten digit recognition needs to improve.Bayesian network due to its graphical model representation, local and distributed learning mechanism, intuitive reasoning; so it is suitable for analysis and expressing uncertainties and probabilistic things; for incomplete, precise or unsure of knowledge and information can make an effective reasoning, and is currently one of the most effective reasoning field model.In order to increase the number identifying the precision of a handwritten, in this paper, Bayesian network and the association rules will be applied in handwritten digit recognition, and proposed a handwritten digit recognition model of Bayesian network. The model is firstly by using binary, thinning, normalized etc pretreatment methods, feature extraction using rough grid to draw eigenvalue; through the network quality standard to measure Bayesian arrangement, choose the practicability of the high quality comparison arrangement method; then, the use of Bayesian network classifier based on association rules related algorithm to distinguish handwritten digits; finally, based on the analysis of experimental results, through the training and testing of data set, test results analyzed and discussed. Because association rules classification is uncertainty, but Bayesian network can overcome this one defect, so use association rules to provide guidance for learning bayesian network will improve bayesian network classifier classification efficiency. The experimental data show that using the method to establish the Bayesian network classification model it has strong classification ability, is a practical and effective identification model. |
PDF Full Text Request