A Machine Learning-Based Calligraphy Font Recognition System Using HOG Features and Support Vector Machine

This study presents an automatic calligraphy font recognition system based on Histogram of Oriented Gradients (HOG) feature extraction and Support Vector Machine (SVM) classification. Addressing challenges such as limited data resources and the complexity of calligraphy scripts, morphological image preprocessing and Canny edge detection techniques were applied for data preprocessing. By integrating HOG feature extraction with SVM classification, the system achieved automatic recognition and classification of five common calligraphy scripts: Seal Script, Clerical Script, Cursive Script, Running Script, and Regular Script. A comparative analysis of SVM models with different kernel functions revealed that the polynomial kernel achieved the best performance with a maximum recognition accuracy of 82.66%. Experimental results demonstrate that the proposed system is highly practical for automatic calligraphy script classification, particularly in scenarios where the script structure is relatively simple and its features are easily extractable.