Res-RNN Network and Its Application in Case Text Recognition

To solve the problem of poor feature extraction ability of traditional text recognition methods in Chinese medical record text, this paper proposes a Res-RNN network for feature extraction based on residual error. Combined with residual characteristics, this network not only improves the depth of the network, but also ensures that there will be no degradation of the network, and strengthens the network's ability to extract Chinese character features. In the residual module, 1 x 1 convolution kernel is used to replace 3 x 3 convolution kernel, effectively reducing the parameters. Combined with feature maps of different scales, the feature information of Chinese characters at different levels is effectively utilized. According to the characteristics of Chinese characters, the vertical sensing field of the feature map is adjusted to retain more vertical fine-grained feature information, thus effectively improving the representational ability of the network. Experiments on actual Chinese medical record text image data set show that the accuracy of the proposed model is 4% higher than that of CRNN.

[1]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2]  Junjun Jiang,et al.  Satellite Image Super-Resolution via Multi-Scale Residual Deep Neural Network , 2019, Remote. Sens..

[3]  Wei Xu,et al.  Bidirectional LSTM-CRF Models for Sequence Tagging , 2015, ArXiv.

[4]  Dong Liu,et al.  Human Face Age Recognition Based on Convolutional Neural Networks , 2018 .

[5]  Johannes Söding,et al.  Protein homology detection by HMM?CHMM comparison , 2005, Bioinform..

[6]  Hossein Ebrahimnezhad,et al.  Three-Dimensional Human Pose Estimation from a Single Image Using Non-linear Convolutional Neural Network Based on Shape Information , 2019 .

[7]  Joelle Pineau,et al.  End-to-End Text Recognition with Hybrid HMM Maxout Models , 2013, ICLR.

[8]  Yaroslav Bulatov,et al.  Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks , 2013, ICLR.

[9]  Wenyu Liu,et al.  Strokelets: A Learned Multi-scale Representation for Scene Text Recognition , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[10]  Farbod Razzazi,et al.  A very high accuracy handwritten character recognition system for Farsi/Arabic digits using Convolutional Neural Networks , 2010, 2010 IEEE Fifth International Conference on Bio-Inspired Computing: Theories and Applications (BIC-TA).

[11]  Jing Wang,et al.  CAPTCHA recognition based on deep convolutional neural network. , 2019, Mathematical biosciences and engineering : MBE.

[12]  Simon Haykin,et al.  GradientBased Learning Applied to Document Recognition , 2001 .

[13]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[14]  Dit-Yan Yeung,et al.  Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting , 2015, NIPS.

[15]  Xiang Bai,et al.  An End-to-End Trainable Neural Network for Image-Based Sequence Recognition and Its Application to Scene Text Recognition , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[16]  Juan Carlos Niebles,et al.  Connectionist Temporal Modeling for Weakly Supervised Action Labeling , 2016, ECCV.