Research on Off-line Handwritten Digit Recognition Algorithm Based on Binary Classification Tree

A handwritten numeric recognition algorithm based on holistic feature (concave and convex) is proposed in document [6], Although the effect is good, this algorithm is only suitable for the sample of the writing specification, in real life, the numbers have a variety of deformations by the randomness of people's writing, it affects the recognition rate of this algorithm. This paper is proposed a novel off-line handwritten numeric recognition algorithm based on binary classification tree. It is improvement through background correction improvement, concave and convex features improvement and classification tree improvement. The experimental results show that the algorithm in this paper has been improved in the original recognition rate and is more adaptive to the deformation of handwritten numerals. The recognition of handwritten numbers is more extensive. At the same time, compared with other popular contour based features and neural network based algorithms, there is an obvious advantage in speed.