Symbiotic Organism Search: A Novel Paradigm for Feature Optimization in Finger Vein Recognition Systems

The main purpose of this work is to build a finger-based biometric authentication system. Hence, the need for high security and efficient authentication approaches, finger vein features are used for person identity. Finger vein is rich in texture information, but it is difficult to extract reliable and accurate features due to poor contrast, lightening, blur, and random noise. Therefore, it is difficult to utilize the complete information of the finger vein features. Information fusion is an essential aspect of biometric systems. Information fusion can occur at various levels of a recognition system. However, feature level fusion is thought to be more effective because a feature set provides more information about the input biometric data than matching score or a classifier’s output decision. Furthermore, in those combined features, there are some mutually exclusive redundant features that will reduce the identification performance. To alleviate this problem, firstly, we propose a new hybrid HOG2LCM multi-algorithm feature extraction technique to extract the texture features from finger vein. Here HOG and GLCM features are integrated at the feature level to extract the complete finger vein texture features. Secondly, to reduce the complexity of feature-level fusion, we proposed a new Symbiotic Organism Search (SOS) optimization technique for selecting features from the Finger Vein. Finally, the K Nearest Neighbor algorithm is used for classification accuracy. The experiment is conducted in THU_FVFDT and FV-USM Datasets finger vein database. Finally, some evaluation metrics are used to assess the performance of our proposed approach. The experimental results show that the proposed method may reduce the number of features from their original size while also increasing the recognition rate. Our proposed solution is implemented using the MATLAB environment.