Multimodal biometrics system: a feature level fusion of physical and behavioral biometric to improve person's identity recognition

Abstract

Multimodal biometric system is the system to prevent a fraudulent attack. The fusion of physical and behavioral modalities can support the needs of the prevention. Thus, we proposed to fuse face and signature as the modalities. In order to get higher performance accuracy, the biometrics is fused at the feature level fusion. LDA is used in order to get the compatible feature vector and in low dimensionality. Multilayer Perceptron is used as the classification technique. The result is evaluated based on GAR, FAR and FRR. The proposed technique able to achieve the GAR of 90%-98.75% depends on the training time and the FAR of below than 10%. Therefore, the proposed algorithm can achieve the main objective of multimodal biometric system and the advantages of the feature level fusion.