Combination of Type III Digit Recognizers using the Dempster-Shafer Theory of Evidence

The Dempster-Shafer Theory of Evidence is an established method for combining different sources of information. In this paper we explore ways to improve the combination performance by building a better BPA for each classifier using both “global” and “local classifier information. We propose modifications to two well-known BPAcomputation methods to make them better suited for combining Type-III classifiers. We also explore the use of compound hypotheses when a classifier cannot confidently choose between the top two returned classes. Experimental tests demonstrate the superiority of some of the approaches proposed here on the numeral recognition problem when combining three different classifiers.

• Type: Conference
• Year: 2003
• Where: ICDAR
• Authors: Catalin I. Tomai, Sargur N. Srihari
• Keywords: Better Bpa, Document Analysis, ICDAR 2003, Local Classifier Information, Well-known Bpacomputation Methods