Gabor Feature Extraction And Recognition For Uyghur Character

With the development of the hand-written recognition technology, this techniquehas already been successfully applied on the recognition of Chinese, English and othercharacters. However, as for Uyghur, this application of this technology is in the initialstage, and there is still a large technological gap needed bridging. In this thesis, in orderto facilitate the application of hand-written character recognition of Uyghur Character,the feature extraction of and the recognition algorithm of hand-written UyghurCharacter are carefully studied. The main contents of this thesis are as follows:1. This thesis introduces the development and the importance of Uyghur characterrecognition, and furthermore several common methods for feature exaction andclassifiers for recognition are described.2. A brief study of the characteristics of Uyghur, which includes the writing style,the connecting forms of Uyghur character, etc. According to the way of writing UyghurCharacter, the difficulties in handwritten Uyghur Character recognition is analyzed andstated.3. This thesis realizes the binarization, smoothing, normalization and Hilditchthinning algorithms to preprocess the handwriting Uyghur Character. And an algorithmfor skew rectification of handwriting Uyghur Character based on Hough Transform isproposed, which could rectify the skew character. Moreover, it could also be applied forthe rectification of skew Uyghur Word.4. An algorithm for energy feature extraction of hand-written Uyghur Characterbased on real value Gabor filter is proposed in this thesis. And the algorithm fordirectional element feature extraction of hand-written Uyghur Character based onuniform grids is realized.5. The recognition of hand-written Uyghur character based on energy featureextraction of real value Gabor filter and KNN classifier is realized. The featureextraction based on directional element features is realized and KNN classifier isdesigned. Furthermore, this thesis combines the two features stated above, and proposesKNN classification algorithm applying both energy features of real value Gabor filter and directional element features. This thesis applies this algorithm to class andrecognize hand-written Uyghur Character, and acquires a good performance. Finally, theanalysis of the result of experiment is presented.