Finger vein authentication using local ternary co-occurrence pattern feature descriptor

This paper proposes a new approach to finger vein authentication using the Local Ternary Co-occurrence Patterns (LTCoP) concept. The illustrated method takes advantage of the unique finger vein patterns in one's finger to accurately authenticate one. The method used is less complex and proves to be highly reliable, offering constructive results for real world applications. The whole research is done keeping in mind the future application of the algorithm on low cost processing devices for practical implementation. The experimental results show that the equal error rate (EER) was 11.8134% and the EER reduces as the number of samples increases proving to be one the most reliable approach for finger vein authentication. The processor used for this project is a low cost 900MHz quad-core ARM Cortex-A7 CPU, which gives a delay of 2.3 seconds for authenticating a single person from 636 classes. This delay reduces as the number of classes reduce.