An improved discrete particle swarm optimization based opposition based learning for palm vein authentication
Lamis Ghoualmi, M.A. Benkechkache, Amer Draa · 2021 International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME) · 2021
Palm vein is among the most secure and accurate biometric technologies. It has got an important interest from the biometric research community. The palm vein patterns represent a physical configuration of a large network of blood vessels located underneath the human skin, which makes it unique. However, the feature extraction stage of the palm veins patterns suffers from multiple disadvantages like a noisy, redundant, and irrelevant data. This kind of feature leads to the decrease in performance of the palm vein biometric system and also leads to the dimension problem. In order to deal with these problems, an improved Discrete Particle Swarm Optimization (DPSO) algorithm based on Opposition Based Learning (OBL) for palm vein authentication is proposed in this work. The proposed DPSO algorithm is used as an optimizer in order to select the best subset of features allowing to enhance the authentication rate of the biometrics system. This Opposition Based Learning (OBL) technique is introduced in the DPSO in order to give more diversity to the population and so to better explore the search space. The method proposed has been tested on the CASIA palm vein database then compared with the traditional based verification system, the Genetic Algorithm (GA), and the DPSO based feature selection. The results obtained show that the method proposed variant outperforms those of the state-of-the-art.