A Reduced Feature Representation Scheme for Offline Handwritten Character Recognition

Recognizing handwritten characters (HCR) is a significant and challenging area of study in the field of computer vision and pattern recognition. Here, a sliding window-based platform for effective visualization of handwritten character images is designed. Three specific feature extraction strategies are suggested to recognize the characters. Those are an angular movement of shape-dependent characteristics, center-to-boundary of the text line, and center-to-boundary of text’s edge feature. The feature selection method is implemented after producing the features to reduce the dimension of the feature matrix. So that, the implementation time can be reduced, somehow. The accuracy from the reduced feature set has also been computed to assess the effectiveness. The feature selection is performed using an extra tree classifier. The characters in this study are recognized using the random forest classifier. The suggested method achieves quite better accuracy with the limited feature set.