Related Experiment Videos
Multimodal biometric system using rank-level fusion approach
Md Maruf Monwar1, Marina L Gavrilova
1Department of Computer Science, University of Calgary, Calgary, AB T2N 1N4 Canada. mmmonwar@cpsc.ucalgary
Summary
Multibiometric systems enhance identity authentication by fusing data from multiple sources, overcoming limitations of single biometric systems. This study demonstrates improved performance using rank-level fusion, even with low-quality data.
Area of Science:
- Computer Science
- Biometrics
- Pattern Recognition
Background:
- Unimodal biometric systems face limitations like noise sensitivity and data quality issues.
- Improving individual biometric matchers often yields limited performance gains.
- Multibiometric systems offer a solution by integrating multiple sources of identity evidence.
Purpose of the Study:
- To present an effective rank-level fusion scheme for multimodal biometric systems.
- To improve identity authentication performance beyond single-biometric capabilities.
- To evaluate the fusion of face, ear, and signature biometrics.
Main Methods:
- Utilized Principal Component Analysis (PCA) and Fisher's Linear Discriminant Analysis (FDA) for individual biometric matchers.
- Developed a novel rank-level fusion method to consolidate results from different biometric modalities.
- Combined matcher ranks using Highest Rank, Borda Count, and Logistic Regression.
Main Results:
- Fusion of individual biometric modalities significantly improved overall system performance.
- The multimodal system demonstrated effectiveness even with low-quality biometric data.
- Rank-level fusion approaches proved successful in consolidating diverse biometric evidence.
Conclusions:
- Multibiometric systems integrated with rank-level fusion offer superior identity authentication.
- The proposed fusion scheme effectively mitigates challenges faced by unimodal systems.
- Rank-level fusion is a viable strategy for designing robust multimodal biometric systems.