Research On Offline Handwritten Chinese Character Cognitive Model And Similar Samples Cognition

For the reason that offline handwritten Chinese character have many specialty including various category of font,high randomness of handwritten,subtle differences between similar samples,offline handwritten Chinese character image machine recognition has became one of the most popular and difficult research on the Pattern recognition field.Considering the traditional open loop recognition model is difficult to satisfy the performance requirements of the off-line handwritten Chinese character recognition,this paper simulates human's thinking information interaction mode with a repeatedly hierarchical comparison from global to local,and explores an feedback intelligent recognition model of offline handwritten Chinese character image with its mechanism in order to improve the cognitive accuracy rate of offline handwritten Chinese character recognition.The main work of this paper is as follows:(1)Building an offline handwritten Chinese character intelligent recognition model with feedback mechanism.This paper optimizes the characterization of discriminating cognitive information between similar samples by establish a feature space with multiple cognitive perspectives in order to accomplish the hierarchical self adjustment of unknown sample's cognitive knowledge space from global to local,and the operation mechanism of the model is given.(2)Research on local discriminating feature extraction method between similar samples.By fusing clustering algorithm and convolutional neural network method,this paper presents an modified convolutional neural network based on clustering algorithm,on the basis of convolutional neural network structure and training algorithm,in order to represent and gain the local discriminating cognitive information between similar samples.(3)Building an recognition results evaluation system of similar samples.On the basis of Latent Semantic Analysis and Information Entropy theory,this paper defines an recognition results evaluation method with calculation model of similar samples to accomplish the self estimation of the intelligent recognition model with feedback mechanism of similar samples and provide the evaluation foundation for the Adaptive optimization adjustment of the recognition results under different cognitive perspectives.(4)Building an offline handwritten Chinese character intelligent recognition system with feedback mechanism.This paper acquires the feature space under multiple cognitive perspectives of offline handwritten Chinese character based on the feature extraction method of offline handwritten Chinese character' global features and similar samples' local discriminating features,and design integrated classifiers to establish the classification rules.To achieve similar off-line handwritten Chinese character recognition,this paper combines the similar sample feedback intelligent identification model and operation mechanism.In order to verify the superiority of the proposed method,this paper uses several samples of GB23122-80 standard simplified handwritten Chinese character library as cognitive object,verified the feasibility and validity of the proposed method by MATLAB simulation experiment.The experimental results show that the average cognitive accuracy rate reached 96.73%,which is better than traditional open-loop cognitive method.