Abstract
In biometrics, periocular recognition analysis is an essential constituent for identifying the human being. Among prevailing the modalities, ocular biometric traits such as iris, sclera and periocular eye movement have experienced noteworthy consciousness in the recent past. In this paper, we are presenting new multi-biometric fusion method called Phase Intensive Mutual Exclusive Distribution (PI-MED) method by combining periocular features (i.e. iris and sclera) for identity verification. The main objective of the proposed PI-MED method is to reduce the matching fusion time and overhead during human recognition in biometrics. Initially, iris modality and sclera modality is pre-processed using Phase Intensive Rubber Sheeting Local Pattern Extraction to generate the vector of score. After that, the extracted iris and sclera features are given to the Mutual Exclusive Bayesian fusion model. The fusion model is applied at the score level for reducing fusion overhead. In this model, feature fusion is generated based on the log likelihood ratio by using covariance matrix measurement. Finally with fusion features, Distributed Hamming Distance Template Matching (DHDTM) algorithm is designed to evaluate the recognition rate of test data with available training data. The results show that the DHDTM significantly improves the recognition rate of human biometric samples when compared to the conventional person identification methods. Several tests were conducted to evaluate the performance of the proposed methods of standard biometric databases using three metrics, namely, matching fusion time, overhead and true positive rate. From the experimental results, the proposed PI-MED method reduces the matching fusion time and overhead by 47% and 45% when compared to existing methods. Similarly, the proposed PI-MED method increases the true positive rate by 33% when compared to existing methods.
Original language | English |
---|---|
Article number | 104024 |
Journal | Image and Vision Computing |
Volume | 103 |
DOIs | |
Publication status | Published - Nov 2020 |
Scopus Subject Areas
- Signal Processing
- Computer Vision and Pattern Recognition
User-Defined Keywords
- Bayesian
- Distributed hamming distance template matching (DHDTM)
- Local pattern extraction
- Mutual exclusive
- Phase intensive
- Rubbersheeting