COMPUTER RECOGNITION OF UNCONSTRAINED HANDWRITTEN NUMERALS

In this paper we present four algorithms developed independently by members of our research team specialized in recognition of unconstrained handwritten numerals. All these methods have high recognition rates and are considered experts by our research group. We also present the different ways experimented on for incorporation of these recognition methods into a more powerful system. By combining them we realize that they complement each other in many ways. The resulting multiple-expert system proves that the consensus of these methods tends to compensate for individual weaknesses, while preserving individual strengths. This paper shows that it is possible to reduce the substitution rate to a desired level while maintaining a fairly high recognition rate in the classification of totally unconstrained handwritten ZIP code numerals. Furthermore, if reliability is of the utmost importance, we can avoid substitutions completely (reliability = 100%) and still retain a recognition rate above 90%. In the last part of this paper, we try to compare results given by some of the most effective numeral recognition systems found in the literature.