A Study Of Smart Off-line Handwritten Chinese Character Recognition System With Feedback Mechanism And Statistical Generalized Characteristics

Offline handwritten Chinese character recognition is a super multi-classpattern recognition problem, which is considered one of the difficult problems inthe field of pattern recognition. At present, open-loop forward is used in themachine identification information mostly and caused the significant contradictionbetween the rate of recognition accuracy and rejection in system performance. Asmart off-line handwritten Chinese character recognition system with feedbackmechanism and the operating mechanism of system is proposed with the imitated ofhuman brain and the intelligent control technology, which had some theoreticalsignificance in promoting the development of the machine identification methodand provided a new way to improve system performance of Chinese characterrecognition.The main works in this paper are as follows:1. A smart off-line handwritten Chinese character recognition system withthree-feedback structure is established and the recognition-output layer, thefeedback layer and the decision-making layer is included in it. The function of eachlayer is defined, and the operating mechanism based on this structure is given toestablish a smart offline handwritten Chinese character recognition system withfeedback correction mechanism.2. Four statistical generalized characteristics errors with random analysis isdefined. Feedback regulation is based on the error correction, so the object to beidentified and identification object are compared to identify four kinds of errors.Rough error of aberration comparison is gotten in statistical generalizedcharacteristic error one. Detail error of character characteristic is gotten instatistical generalized characteristic error two. Basis for qualitative andquantitative is provided for the system by the two errors. Characteristic of detailerror is gotten in statistical generalized characteristic error three. Matching trenderror is gotten in statistical generalized characteristic error four. Basis for feedbackcorrection mechanism is provided for the system by the two errors.3. Two literacy mechanisms are established for feedback correction. Matchingtrend error is used to the regulate feedback and judge the effect ofrecognition of one wheel by imitating the human brain. The smart method of multi-recognizer integration is used. Characteristic of detail error is used to adjustthe feedback loop and select the recognizer of new loop. Operating mechanism ismade more reasonable by feedback control ideology and intelligent controltechnology.Chinese sample characters in SCUT-IRAC HCCLIB are used to simulation bythe system. VC++and MATLAB7.3simulation software are used and400handwritten Chinese characters in font are selected in experiment to validate thesmart handwritten Chinese character recognition system with feedback mechanismand statistical generalized characteristics build in this paper. Experimental resultsshow that the method is feasible and effective.