Offline Handwritten Chinese Character Recognition Based On Cascade Classifier

Chinese character recognition belongs to the text recognition,and text recognition is one of the important research fields of pattern recognition.Chinese character recognition involves digital image processing,pattern recognition,statistics,natural language processing,information theory and other disciplines.There is less available information for offline handwritten Chinese character recognition,and the recognition is relatively difficult.The main contents of this thesis include:(1)Offline handwritten Chinese character feature extraction method.Feature extraction plays a crucial role in offline handwritten Chinese character recognition.An excellent feature extraction method should be able to deal with such issues as stroke position offset and tilt.This thesis analyzes the elastic grid feature extraction method in offline handwritten Chinese character recognition.The grid method ignores the problem of the connection between grids.Gaussian distribution weighting template is introduced into the original elastic grid method,and the validity of the method is illustrated by the evaluation standard of variance ratio between classes.(2)Similar Chinese characters.The problem of similar Chinese characters is one of the difficulties in offline handwritten Chinese character recognition.There are a large number of similar Chinese characters.The composition difference of many characters is always small.And this makes the offline handwritten Chinese character recognition challenging.This thesis analyzes and summarizes the characteristics of similar Chinese characters.Through the study of the modified quadratic discriminant function(MQDF)theory,a method of obtaining similar Chinese character set through the MQDF is proposed.(3)Cascade classifier.Classifier is the most important part of offline handwritten Chinese character recognition.In this thesis,the support vector machine and modified quadratic discriminant function is used to identify all the similar pairs of Chinese characters and establish a set of similar Chinese character recognition models.Then,using MQDF classifier as the first classifier and similar Chinese character recognition model set as the second classifier,MQDF-SVM and MQDF-MQDF cascade classifier model is established to classify offline handwritten Chinese characters.