Research On The Evaluation Method Of Hard Pen Chinese Character Writing Quality Based On Template Matching

Under the information age,the ability of Chinese people's hand-writing is getting worse in recent decades.In this very case,the education of Chinese character writing skill has drawn widespread attention from schools,educational departments and all over the society since it is deemed as an essential and basic skill of people.To put it into practice,one of the most serious dilemmas for the education of Chinese character writing skill is the lack of professional teachers and specialized guidance in actual teaching environment.Starting from template-based method for Chinese character writing evaluation,this paper,therefore,attempts to investigate a set of feasible evaluation plans for hard-pen Chinese character writing.The major issues to be addressed in the paper are as follows:The digitalized pre-processing of Chinese character hand writing information is the premise and foundation of follow-up evaluation.Based on the android mobile data collection platform,this paper adopts XML,using various methods like smoothing-denoising,passivation-denoising,stroke segment splitting and combination,distractive stroke segment removing and stroke completion to do digitalized pretreatment of hand writing information which prove to be effective.This paper evaluates the quality of hard-pen Chinese character writing from the perspectives of both correctness and neatness.Correctness evaluation starts with the analysis of relevant features of Chinese characters(e.g.,the type,number and order of strokes,the spatial relationship of similar strokes,the length and topological relationship of different strokes,and the spatial position of stroke units).These features are extracted and compared with model Chinese characters for correctness evaluation.Results showed that,for all the aforementioned seven indexes,the consistency rate of expert evaluation and machine evaluation exceeded 98%,which illustrated the quality evaluation of this research.Concerning neatness evaluation,this paper put forward two methods which are based on feature labeling and "Mass-Spring" energy model.The feature-label-based method extracts features of the written Chinese characters(e.g.,the alignment and length proportional relationship of similar strokes,joint points of different strokes,the gradient of strokes,the alignment,density,size relationship and aspect ratio of stroke units,the position,size and aspect ratio of entire character)and compares them with standard Chinese characters so as to obtain the similarity degree of respective features,thereby working out the evaluation of sub-items.The overall feather similarity is calculated by giving feature weight(adjusted by BP neural network)to sub-items,and through synthetic computation,the neatness evaluation is completed.The "Mass-Spring" energy model based method adopts elastic meshing.The strokes,stroke units and the entire written Chinese character are compared with standard Chinese character in terms of elastic meshing deflection and its resultant energy difference.Results indicated that,for the feature-label-based method,when redundancy value ?= 5,the consistency rate of expert evaluation and machine evaluation was 80.24%,when redundancy value?= 10,this rate reached 93.21%.As for the "mass-spring" energy model based method,the consistency rates were 78.57%and 92.14%respectively,which were lower than feature-label-based method on the whole.Based on the research metioned above,we also designed a real-time and template-based scheme for evaluating the quality of hard-pen Chinese character writing.This scheme implements a number of algorithms,including data pre-processing,correctness evaluation,neatness evaluation,evaluation and guidance generation.The prototype system employs Eclipse+Android and object-oriented architecture,and the backend database adopts SQL Sever2012,all of which prove to be superior in robustness and expansibility.