Correlation coefficient based template matching: Accounting for uncertainty in selecting the winner

The problem of selecting a template that matches a given candidate signal is applicable across a wide variety of domains. Using the correlation coefficient as the avenue for selecting the winning template is perhaps the most common technique. The challenge lies in selecting the winning template when there is no clear separation between the correlation coefficient values of the winning template and the others. In this paper, we present a simple Dempster-Shafer (DS) theoretic model that enables one to capture the uncertainty regarding the winner selection in correlation coefficient based template matching. The DS theoretic framework provides an avenue to develop the model with few resources and little to no prior knowledge. We validate the model using several numerical examples and a numerical character recognition application where the evidence provided by several sets of templates are combined using a DS theoretic fusion strategy to arrive at a better decision.