Skip to Main Content
This paper describes the detailed process required to accomplish the printed Yi optical character recognition, including pretreatment, feature extraction and pattern classification. The arithmetic of Yi multifontpsilas feature extraction, the arithmetic of recognition required for dictionary building and the arithmetic of multilevel Yi character matching were analyzed with particular emphases. An OCR software based on Windows with self-correcting ability was developed. Its one-time accurate identification of samples was above 99.5%. This result proved that the arithmetic used is feasible and efficient.