Mutibiometric System Based on Game Theory

Biometric systems based on a single biometric trait have drawbacks that are alleviated by multibiometric systems, which combine multiple sources of information. The novelty of this research effort is that Coalition Game Theory (CGT) is applied to improve the performance of the iris and face based multibiometric system. The CGT technique selects the most salient patches obtained using the Local Binary Patterns (LBP) and modified LBP (mLBP) feature extraction techniques. The CGT chooses patches that have better individual importance along with a strong interaction with other patches based on the Shapely value. Results show that CGT model maintains impressive recognition accuracy while using smaller image areas for recognition. More specifically, CGT outperforms the LBP and mLBP techniques.