FINGER VEIN RECOGNITION USING ROBUST FEATURE EXTRACTION AND SVM
| Main Author: | Kanika Kapoor*1 & Vinay Chopra2 |
|---|---|
| Format: | Article eJournal |
| Terbitan: |
, 2019
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/3235805 |
Daftar Isi:
- In today’s society, the identity verification is a serious key problem with the rapid development in the domain of computer and network technology. Hence, the necessity for a superior and more consistent methodology for identity authentication becomes more prominent. As biometric identifiers are relatively tough to counterfeit, mislay or share, biometric recognition approach appears to be more effective and reliable than conventional passwords or PINs. Owing to its low forgery risk, consistency and aliveness detection, Finger Vein Recognition (FVR) has emerged to be the most promising and novelist biometric technique. Finger vein pattern is defined as the hypodermic vein structures arbitrarily developing a network of blood vessels underneath the skin of a finger to recognize individuals at a very high level of accuracy. However, it is challenging to extract a more reliable and accurate finger vein pattern due to the random noise, low contrast, illumination variation, image deformation and blur. Not much research has been conducted on effective frequency domain feature extraction techniques, hence, considering the above issues, this research presents an efficient feature extraction approach which employs the Local Directional Pattern (LDP), which is robust in the existence of random noise, ageing effects as well as illumination changes. Support Vector machines (SVM), which is a powerful machine-learning binary classifier, is implemented in order to enhance the recognition performance by classifying finger vein patterns as either imposter or genuine. The experimental results demonstrate that the proposed approach achieved significant performance and better classification accuracy on HKPU database. An accuracy of 97.5% with an Equal Error Rate of 0.81% is achieved indicating superior results over existing techniques.