User-Independent Online Handwritten Digit Recognition

This paper proposes a fast user-independent method for handwritten digit recognition. The local feature of inputting strokes is firstly coded according to the eight equiangular encircled directions. Inputting digit is then modeled with a set of rules defined with the code of local features to characterize the drawing style of inputting digit. The decision tree learning is also invoked to model the variance of drawing styles and guarantees high recognition rate. Main advantage of proposed method is twofold. Firstly, it is quite simple and highly discriminating, and can do recognition quickly under strict resource constraints. Secondly, it is insensitive to different users and guarantees user adaptability. Experiments prove our method both effective and efficient for online handwriting digit recognition