基于HOG特征提取和支持向量机的东巴文识别

Dongba script，created by Naxi minority in China，is the only surviving pictograph in the world，facing difficulties in inheritance currently. The automatic recognition of Dongba script through machine learning has been studied. By scanning，Dongba script is converted into digital image，and then the grayscale image is smoothed. At last，the contour of Dongba script is extracted with the edge detection algorithm. The feature vectors extracted by histogram of oriented gradient (HOG) are input into support vector machine (SVM) for the image classification so as to recognize Dongba script automatically. Through the test of self built data set，under the same experimental conditions，the recognition rate of SVM is higher than that of K⁃nearest neighbor classifier and random forest classifier，reaching 97.42%，and the recognition process is simple and fast. It has importantly practical significance for the inheritance and protection of Dongba script.

Keywords： Dongba script ; feature extraction ; histogram of oriented gradient (HOG) ; Support Vector Machine (SVM) ; character recognition