Semester

Summer

Date of Graduation

2010

Document Type

Thesis

Degree Type

MS

College

Eberly College of Arts and Sciences

Department

Economics

Committee Chair

James Harner

Committee Co-Chair

Arun Ross

Abstract

Biometrics refers to the automatic recognition of individuals based on their physical or behavioral characteristics. Multimodal biometric systems, which consolidate multiple biometric characteristics of the same person, can overcome several practical problems that occur in single modality biometric systems. While fusion can be accomplished at various levels in a multimodal biometric system, score level fusion is commonly used as it offers a good trade-off between fusion complexity and data availability. However, missing scores affect the implementation of most biometric fusion rules. While there are several techniques for handling missing data, the imputation scheme, which replaces missing values with predicted values, is preferred since this scheme can be followed by a standard fusion scheme designed for complete data. Performance of the following imputation methods are compared: Mean/Median Imputation, K-Nearest Neighbor (KNN) Imputation and Imputation via Maximum Likelihood Estimation (MLE). A novel imputation method based on Gaussian Mixture Model (GMM) assumption is also introduced and it exhibits markedly better fusion performance than the other methods because of its ability to preserve the local structure of the score distribution. Experiments on the MSU database assess the robustness of the schemes in handling missing scores at different training set sizes and various missing rates.

Share

COinS