GB1359781A - Alpha-numeric character recognition system - Google Patents

Abstract

1359781 Character recognition BENDIX CORP 3 Jan 1972 [7 Jan 1971] 156/72 Heading G4R In a character recognition system binary signals representing spaced groups of horizontal sections and spaced groups of vertical sections of the scanned character are examined. A bit is generated for each group dependent upon whether or not the section having the most 1- bits in the group exceeds a certain adjustable threshold. The bits generated identify the character despite skew or off-centring. As disclosed the characters move past a scanner which scans four characters at a time. Character movement is frozen by a shutter associated with the scanner. The binary signals representing the four characters are stored, Sections of each character in turn are now passed to a processor. Consider for example the skewed character J in Fig. 6. Eight groups of sections are taken, i.e. groups 1-3 of vertical sections 1 1 , 12, 1 3 , 2 1 , 2 2 ; &c., and groups 4-8 of horizontal sections, as shown. The number of 1-bits in the section having the greatest number of 1-bits in a group, e.g. section 5 1 in group 5, is compared with a threshold number, e.g. 12, and if it exceeds this threshold a 1-bit is generated. The 8-bit signal resulting from this operation on the 8 groups is used to identify the character, e.g. by means of a tree search. If the character is identified as being one of a set of characters whose 8-bit identifying signals are closely similar the groups of sections are taken again but this time are compared with an increased threshold, e.g. 18. The identity of the character may be shown on a display or be printed. Reference is made to locating marks which accompany the characters scanned and to segmentation of characters.