Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Jeremy M Dawson

Committee Co-Chair

Bojan Cukic

Committee Member

Matthew Valenti


Biometric data is an essential feature employed in testing the performance of any real time biometric recognition system prior to its usage. The variations introduced in the match performance critically determine the authenticity of the biometric data to be able to be used in an everyday scenario for the testing of biometric verification systems. This study in totality aims at understanding the impact of data stratification of a such a biometric test dataset on the match performance of each of its stratum. In order to achieve this goal, the fingerprint dataset of the West Virginia University's 2012 BioCOP has been employed which is a part of the many multimodal biometric data collection projects that the University has accomplished. This test dataset has been initially segmented based on the scanners employed in the process of data acquisition to check for the variations in match performance with reference to the acquisition device. The secondary stage of data stratification included the creation of stratum based on the demographic features of the subjects in the dataset.;The main objectives this study aims to achieve are:;• Developing a framework to assess the match score distributions of each stratum..;• Assessing the match performance of demographic strata in comparison to the total dataset..;• Statistical match performance evaluation using match score statistics..;Following the generation of genuine and imposter match score distributions , Receiver Operating Characteristic Curves (ROC) were plotted to compare the match performance of each demographic stratum with respect to the total dataset. The divergence measures KLD and JSD have been calculated which signify the amount of variation between the match score distributions of each stratum. With the help of these procedures, the task of estimating the effect of data stratification on the match performance has been accomplished which serves as a measure of understanding the impact of this fingerprint dataset when used for biometric testing purposes.