Techniques for Highly Accurate Optical Recognition of Handwritten Characters and Their Application to Sixth Chinese National Population Census

Highly accurate optical character recognition (OCR) of handwritten characters is still a challenging task, especially for languages like Chinese and Japanese. To improve the accuracy, we developed four techniques for enhanced recognition: character recognition based on modified linear discriminant analysis (MLDA), subspace-based similar-character discrimination, multi-classifier combination, and mutual-information-based adaptive rejection. They were applied by the Chinese government to the Sixth National Population Census in 2010. By combining address and nationality information, they achieved an accuracy of over 99% with a low rejection rate. This was the first time that optical recognition of handwritten Chinese characters had been used on a large-scale in the Chinese census project.