论文信息 - Online Handwriting Mongolia Words Recognition with Recurrent Neural Networks

Online Handwriting Mongolia Words Recognition with Recurrent Neural Networks

This paper primarily discussed Online Handwriting Recognition methods for Mongolia words which being often used among the Mongolia people in the North China. Because of the characteristic of the whole body of the Mongolia words, namely connectivity between the characters, thereby the segmentation of Mongolia words is very difficult. We introduced a recurrent neural network to online handwriting Mongolia words recognition. The system consists of an advanced recurrent neural network with an output layer designed for sequence labeling, partially combined with a probabilistic language model. Experimental results show that unconstrained Mongolia words achieve recognition rates about 80%, compared with about 70% using a previous developed HMM-based recognition system.

Wu Wei | Gao Guanglai

[1] Jürgen Schmidhuber,et al. Framewise phoneme classification with bidirectional LSTM and other neural network architectures , 2005, Neural Networks.

[2] Yoshua Bengio,et al. Gradient Flow in Recurrent Nets: the Difficulty of Learning Long-Term Dependencies , 2001 .

[3] John Scott Bridle,et al. Probabilistic Interpretation of Feedforward Classification Network Outputs, with Relationships to Statistical Pattern Recognition , 1989, NATO Neurocomputing.

[4] Kuldip K. Paliwal,et al. Bidirectional recurrent neural networks , 1997, IEEE Trans. Signal Process..

[5] Jürgen Schmidhuber,et al. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks , 2006, ICML.

[6] Jürgen Schmidhuber,et al. Long Short-Term Memory , 1997, Neural Computation.

[7] Lawrence R. Rabiner,et al. A tutorial on hidden Markov models and selected applications in speech recognition , 1989, Proc. IEEE.

[8] Homayoon Sadr Mohammed Beigi,et al. Pre-Processing The Dynamics Of On-Line Handwriting Data, Feature Extraction And Recognition , 1996 .