Research On Android Platform Based Kazakh Slavonic Recognition System

With the rapid development of computer technology and information processing technology, character recognition being as the main component part in pattern recognition and multi-language information processing system, it takes an indispensable role in the field of pattern recognition and image processing. Android system is a new developed operation system of mobile terminal, and it has a strong network application and system function.in recent years, Android mobile terminal being as an universally applied communication and information tool,application range became wider,number of users were increasingly risen.compare with computer and Internet, Android mobile terminal can provide people with more immediate and more convenient operating environment. With the improvement of optical character recognition technology, relevant script classes and operation platform of character recognition diversify, combination of mobile terminal and document image character recognition emerged as new developed research area,and Android mobile terminal platform based multi-language character recognition system have developed gradually.In order to develop the using of Android mobile phones in pattern recognition,and recognize the Kazakh Slavic characters accurately by mobile phone, Android platform based Kazakh Slavic optical character recognition system is proposed in this paper. Firstly, pre-processing is made to the document images in the mobile phone;image objects are segmented into words by horizontal segmentation and obtained characters by vertical segmentation; connected characters are segmented by over segmentation. And then, divided character images are normalized into 27 ? 27 points,make rough classification based on X-axis symmetric characteristics, Y-axissymmetric characteristics, and both X-axis and Y-axis symmetric characteristic. Three different through features includes horizontal vector, vertical vector and diagonal vector features are extracted, then based on this features make second step classification; Finally, the character images are normalized into 9? 9 points again, and81 dimension Grid features are extracted, then by training and classification, satisfied recognition results are showed. Differ from the traditional feature sets, a tree XML file feature set is established in this experiment. Compared with the earlier proposed methods, our method is more efficient with the fast running speed and higher recognition rate.