Fusion of face and signature at the feature level by using correlation pattern recognition

Abstract

A combination of more than one biometric is the enhancement of unimodal biometric. It is called multimodal biometric. A feature level fusion is one the fusion level. To date, feature level fusion is less implemented due to the difficulties in combining the feature from different modalities. We combined the feature of face and signature that from the different domain. Correlation pattern recognition with MACE filter is used to overcome the problem of different domain. By using MACE filter, we are able to extract the feature from face and signature and produce a new fused feature vector in a frequency domain. We used a threshold specification to identify the sample testing that genuine or impostor. The Genuine Acceptance Rate (GAR) and False Acceptance Rate (FAR) are the component to evaluate the system performance. The proposed work is able to achieve preliminary GAR of 85.71 % and FAR of 14.29%-20%. Keywords—multimodal biometrics, feature level fusion, correlation pattern recognition. B I.