SEJONG-NET with analysis-by-synthesis

SEJONG-NET, inspired by the visual system of vertebrate animals, has been designed to recognize dynamic characters, and applied to various character sets such as Hangul (Korean alphabet), Roman alphabet, Chinese character and numerals. In this paper, we propose the extended SEJONG-NET model based on the generalized model for dynamic character recognition. This model can recognize both dynamic patterns and static patterns through a common mechanism of recognition. In this model we treat the process of static character recognition as a special case of dynamic character recognition through selectively scanning over strokes according to the remembered writing sequence.