Abstract:
Objectives In order to solve the problem that the accuracy of handwritten Chinese character text recognition is not high, an end-to-end method for handwritten Chinese character text recognition based on convolutional neural network and recurrent neural network is proposed.
Methods Firstly, the convolutional neural network constructed by inception module is used to extract the basic features of the text image. Secondly, the recurrent neural network is used to predict the extracted features and output a probability distribution about the Chinese character set. Finally, the connectionist temporal classification algorithm is used to calculate the recognition results and construct the loss function.
Results The proposed method is tested on the handwritten Chinese character text dataset, experimental result shows that the Inception module and data enhancement method can effectively improve the performance of the algorithm, obtain the recognition accuracy of 71.2% and the text editing distance of 0.060.
Conclusion Our proposed method can conduct end-to-end handwritten Chinese character text recognition, and improve the recognition accuracy compared with the existing methods.
下载:
