Research On Handwritten Digits Recognition Method Combining Multivariate Coding And Rule Evaluation

Handwriting character recognition has been a research hotspot in the field of pattern recognition and image processing in recent years.Although the problem is simple,it has great practical value.Aiming at the problem of low accuracy of existing character image recognition algorithms,this paper provides a simple recognition method which is easy to understand and process.This recognition method can be used in common handwriting recognition systems,and in addition to Arabic Digits,it can process variety of other Characters as well.The specific idea is: using character image feature extraction algorithm to extract features.The main idea of feature extraction algorithm is to extract the 0-pixel points in the binarized character image,surrounded by 1-pixel points in several directions,then save the number of directions as features.And then construct two different KNN classifiers based on the features to classify the characters.Because there may be some samples of classification errors in the classification results of the two classifiers,according to the possible errors,the corresponding fusion rule set is formulated to correct the classification results and further improve the accuracy of character recognition.The experimental results on UCI datasets show that the algorithm is superior to the recognition algorithm based on single KNN classification model.The above research broadens the research scope of feature extraction algorithm,provides new research direction for the design of consistent feature extraction algorithm,and also lays a foundation for the research of feature extraction.It is believed that the continuous research of such algorithms can solve more practical problems.