Abstract
GABOR FILTER-BASED FEATURE LEVEL FUSION OF PALM VEIN AND FINGERPRINT RECOGNITION SYSTEM
*Ganiyu Rafiu Adesina, Omidiora Elijah Olusayo, Olayiwola Dare Samseedeen
ABSTRACT
Biometrics fusion entails using two or more physiological or behavioral traits to improve the performance of biometric systems. Most existing works investigated effects of fusion of multiple features at image, matching score and decision levels for biometric recognition systems. In this paper, an efficient Gabor Filter-based Feature Level (GFFL) fusion of palm vein and fingerprint recognition system was developed. Four hundred images consisting of five palm vein and fingerprint images each across 40 individuals were acquired. Two hundred and ten images were used for training while the remaining 190 images were used for testing purposes. These images were preprocessed using histogram equalization while the extraction of features from the region of interest of the images was carried out using Gabor filter algorithm. The extracted features were fused through concatenation and subjected to Fisher Linear Discriminant Analysis for dimensionality reduction. Euclidean distance was used in the classification of the images. The system was implemented using Matrix Laboratory 7.1. The performance of the developed feature level fusion system was evaluated at varying thresholds (0.2 - 0.8) by comparing it with matching score level fusion biometric system as well as existing Fingerprint and Palm vein unimodal systems based on recognition accuracy, sensitivity, specificity and false positive rate. The evaluation results revealed that the developed GFFL fusion of palm vein and fingerprint recognition system outperformed the existing unimodal systems in terms of the aforementioned metrics. The system could be adopted as effective recognition system in access control systems or any other related systems.
[Full Text Article] [Download Certificate]