Vietnamese handwriting recognition for automatic data entry in enrollment forms

摘要

This paper presents an efficient method of feature extraction for Vietnamese handwriting and an approach to group Vietnamese characters into similar sub-classes to support recognition process. Furthermore, it's the algorithm to look up two dictionaries simultaneously with least database storage in the post-processing in order to increase the recognition accuracy. The results of our approach on recognizing Vietnamese handwriting such as name, birthplace and birthday are also presented. By using Vietnamese Personal Name Dictionary for post-processing, the recognition accuracy for Vietnamese characters, words, person name is 91.23%, 85.41% and 96.03% in comparison with 75% 79.67% and 77.81% from common methods correlatively. The statistics are based on 2384, 2248 and 302 testing samples of Vietnamese character, words and personal name respectively. The recognition accuracy for province/city name is 99.8% among 64 provinces of Vietnam. Our approach is embedded into the automatic data entry system for enrollment forms, VnHandwritten1.0, the first commercial application in recognizing Vietnamese handwriting in Vietnam. On this application, our method proves itself an efficient method for commercial purpose.