Academic Editor: Youssef EL FOUTAYENI
Received |
Accepted |
Published |
01 February 2020 |
16 February 2020 |
10 March 2020 |
Abstract: Biometrics is a field that is undergoing rapid change on everything with technological development and acquisition devices. Biometrics significantly improves the security of information. The mono-modal systems have brought their points of satisfaction, but on the other hand they presented a disadvantage, which is the intrusion and the rate of identification. Researchers and manufacturers in the field have opted for a multimodal biometric recognition system. They make it possible to improve the score such as the accuracy and the rate of people recognition (fusion of descriptors between several biometric methods). In our article we propose to develop a multimodal biometric system based on the face and fingerprints. Facial recognition uses several mathematical methods to extract the relevant information from the face which allows the descriptor [1] [2] to be extracted, among these methods we find: global methods (ACP [3], LDA [4]), local methods (SVM [5], HMM [6]), hybrid methods [7] [8]. Fingerprints recognition is delicate because it goes through several processes to be able to extract the descriptor (binarization, skeletonization, detection of minutiae, elimination of false minutiae) [9] [10]. In our approach we recommend using Deep learning technology with convolutional neural networks. We will retrieve the descriptors of the face and the fingerprints, then we use two CNN architects namely VGG 16 [11] and Resnet 50 [12]. The experiments were carried out on the SDULMA-HCM database [13].
