An Algorithm of Bidirectional RNN for Offline Handwritten Chinese Text Recognition.

Handwritten Chinese text recognition characters is a challenging problem as it involves a imbalanced training data, and the samples are very different even in same character. In this paper, we propose a novel algorithm based on the bidirectional Recurrent Neural Network (BiRNN) to recognize the characters in the text regions. We solve the problems with pre-processing and improved CNN network. In addition, we utilize RNN to analyze the correlation between characters. Compared with previous works, the algorithm has three distinctive properties: (1) It can predict characters by context analyzing from forward and backward. (2) It solve the problem of sample imbalance effectively. (3) The convergence rate of training has increased. Moreover, the proposed algorithm has achieved good results in recognition.