SEJONG-NET is a neural network model for providing a proper paradign the process of visual character pattern recognition.In this paper,we propose the SEJONG-NET model which can recognize both on-line and off-line Hangul characters with the common recotgnition mechanism.The extended SEJON-NET,based on the on-line Hangul model,have various backward processes to analyze and synthesize the information on strokes when the model fails to find out the valid outputs in the higher stages.Through the backward processes using the hierarchical structural knowledge of Hangul,the model can recover the temporal information from off-line characters.Once the dynamic information is identified,we consider the off-line character as the on-line one,and recognized it by means of the on-line mechanism. The simulation of the proposed model showed that it recognized the off-line characters through recivering the temporal informations as well on-line characters with the common mechanism of recognition.The restored dynamic information was almost identical to the human writing sequence.Also,using the backward paths, the model could recognize on-line characters with the variations of strokes.Hence we may consider that the extened model has the better recognition mechanism and is a more cognitively plausible model of problem solving for the tasks of Hangul character recognition. Keywords:Hangul Recognition,Neural Network Model,Recovery of Temporal Imformation,Common Mechanism for On-line/ off-line Recognition.