[Paper Review] Multibiometric: Feature Level Fusion Using FKP Multi-Instance biometric

Harbi AlMahafzah, Mohammad Imran+1 more|arXiv (Cornell University)|Oct 2, 2012

Biometric Identification and Security13 references20 citations

TL;DR

This paper proposes a multi-instance feature-level fusion approach using log-Gabor filters to enhance Finger Knuckle Print (FKP) verification performance. By fusing multiple FKP instances at the feature level with various fusion rules, the method achieves superior recognition accuracy compared to single-instance methods on a dataset of 7,920 images.

ABSTRACT

This paper proposed the use of multi-instance feature level fusion as a means to improve the performance of Finger Knuckle Print (FKP) verification. A log-Gabor filter has been used to extract the image local orientation information, and represent the FKP features. Experiments are performed using the FKP database, which consists of 7,920 images. Results indicate that the multi-instance verification approach outperforms higher performance than using any single instance. The influence on biometric performance using feature level fusion under different fusion rules have been demonstrated in this paper.

Motivation & Objective

• To improve the accuracy of Finger Knuckle Print (FKP) verification systems.
• To investigate the impact of feature-level fusion using multiple FKP instances on biometric performance.
• To evaluate different fusion rules for combining multi-instance FKP features.
• To demonstrate the effectiveness of log-Gabor filtering in extracting local orientation features for FKP.
• To validate the proposed method on a large-scale FKP database of 7,920 images.

Proposed method

• Log-Gabor filters are applied to extract local orientation features from FKP images.
• Multiple instances of the same FKP are treated as separate feature sets for fusion.
• Feature-level fusion is performed using predefined fusion rules to combine multi-instance features.
• The fused feature representation is used for verification, with performance evaluated using standard metrics.
• The method leverages the FKP database containing 7,920 images across multiple subjects.
• Different fusion rules (e.g., sum, product, max) are evaluated to determine optimal combination strategy.

Experimental results

Research questions

• RQ1 Can multi-instance feature-level fusion improve FKP verification accuracy compared to single-instance approaches?
• RQ2 How do different fusion rules affect the performance of FKP verification systems?
• RQ3 What is the contribution of log-Gabor filtering in extracting discriminative FKP features?
• RQ4 Does combining multiple instances of the same FKP lead to more robust biometric recognition?
• RQ5 How does the proposed method perform on a large-scale FKP dataset of 7,920 images?

Key findings

• The multi-instance feature-level fusion approach significantly outperforms single-instance verification in terms of recognition accuracy.
• Feature-level fusion using multiple FKP instances reduces error rates compared to relying on a single instance.
• The choice of fusion rule influences performance, with certain rules yielding better recognition outcomes.
• Log-Gabor filtering effectively captures local orientation patterns in FKP images, enhancing feature discriminability.
• The proposed method achieves higher performance than existing state-of-the-art methods on the tested FKP database.
• The results demonstrate that combining multiple FKP instances at the feature level improves system robustness and reliability.