JPS57113184A - Character recognizing method - Google Patents

Abstract

PURPOSE:To improve the precision of recognition by extracting and cnverting the features of a character into a binary signal, reading and converting a dictionary, arrayed by the features, and a handwritten character into binary signals, and providing a weighting mask to a similar character part while comparing other parts with said dictionary. CONSTITUTION:An image of a handwritten character 3 is picked up by a vidicon 5, whose output is converted photoelectrically; and a segmenting circuit 6 segments the character, and a feature extracting circuit 8 extracts and classifies its features through a normalizing circuit 7. Consequently, sequential matching 9 with binary-coded characters stored in a dictionary 10 previously is performed, and thus recognition 11 is performed by using the dictionary 10. Prospective characters selected in the dictionary 10 are masked by a mask processing circuit 12 and the 2nd recognition 13 is performed. For example, three sets of (AB), (AC) and (AD) having a left hand radical (A) have right-hand radicals (B), (C) and (D) which are obviously different while having less difference in left- hand radical (A). Those left-hand radical parts are mesked and only the right- hand radicals are matched. Thus, the precision of recognition is improved while a recognition rate is increased.