Multiple combined recognition system for automatic processing of credit card slip applications

The authors developed a system that utilizes a multiple combined method. For the recognizer, they used three representative recognizers: a structural, a statistical, and a neural network approach. They also used three combining methods (a vote, a Bayesian, and BKS) to combine the results obtained from three recognizers. This system is applied to credit card slip recognition. For recognizing printed numerals, the authors use template matching, and for handwritten numerals, a multiple combined system is used. Reliability about 99% was achieved in handwritten numeral recognition and 81% reliability in credit card slip recognition. L. Lam and C. Y. Suen´s (1988) standard handwritten numeral data showing high reliability of above 95%