A new algorithm for handwritten character recognition
Zhu Xiaoyan, Shi Yifan · 2002
A new algorithm of handwritten character recognition based on feedback theory is proposed. We suggest this new method by adding confidence back-propagation and input modification on the neural network model, thus preprocessing and recognition are integrated closely. Convergence of the algorithm is proved. Experiments show that it greatly reduced the system's error rate and was robust to environmental noise.