Research Article
Handwritten Uyghur Character Recognition Using Convolutional Neural Networks
@INPROCEEDINGS{10.1007/978-3-030-72795-6_49, author={Wujiahemaiti Simayi and Mayire Ibrayim and Askar Hamdulla}, title={Handwritten Uyghur Character Recognition Using Convolutional Neural Networks}, proceedings={Simulation Tools and Techniques. 12th EAI International Conference, SIMUtools 2020, Guiyang, China, August 28-29, 2020, Proceedings, Part II}, proceedings_a={SIMUTOOLS PART 2}, year={2021}, month={4}, keywords={Handwritten character recognition Uyghur character forms Size adjusting Convolutional neural network}, doi={10.1007/978-3-030-72795-6_49} }- Wujiahemaiti Simayi
Mayire Ibrayim
Askar Hamdulla
Year: 2021
Handwritten Uyghur Character Recognition Using Convolutional Neural Networks
SIMUTOOLS PART 2
Springer
DOI: 10.1007/978-3-030-72795-6_49
Abstract
Handwritten Uyghur character recognition researches up to date have been based on traditional pattern recognition techniques that highly relies on handcrafted features. The similarity between character forms has been hindering the extraction of robust features. This paper proposed the deep learning based self-learned features to recognize 128 handwritten Uyghur characters forms. The first-hand online handwritten trajectory is first preprocessed and converted to a centralized binary image as input to the implemented deep neural network model. In experiments, the convolutional neural network models with 4, 5 and 8 convolutional layers are studied to get higher recognition accuracy. All models are trained implementing the same dropout regularization. The models 8 convolutional layers on 48 * 48 converted character images produced as high as 94.65% average accuracy on a test set of 10,240 handwritten character samples.
