A Multi-modal Biometric Approach Based on Score-level Fusion and Fine-tuning Deep Learning Features

Alghamdi, Mona (2023) A Multi-modal Biometric Approach Based on Score-level Fusion and Fine-tuning Deep Learning Features. In: 2022 IEEE 11th International Conference on Intelligent Systems, IS 2022 :. IEEE, POL, pp. 1-6. ISBN 9781665492768

[thumbnail of Fusion_conference_paper_pure_version]
Text (Fusion_conference_paper_pure_version)
Fusion_conference_paper_pure_version.pdf - Accepted Version
Available under License Creative Commons Attribution.

Download (310kB)

Abstract

This paper presents a multimodal biometric approach applied to all fingernails and knuckle creases of the five human fingers for identifying persons. In this paper, the proposed biometric technique consists of several phases. The method starts with the detection and localisation of the main components of the hand, defining the region of interest (ROI), segmentation, feature extraction by retraining the DenseNet201 model, measuring the similarity using different metrics, and lastly, improving the person identification performance by implementing score-level fusion. This approach presents different methods for person identification, which combine fingernails, knuckles based on the modality type, and whole hands based on different similarity metrics. This paper uses various similarity metrics to distinguish between individuals. These include the Bray-Curtis, Cosine, and Euclidean metrics. Two main score-level fusion techniques are employed: the majority voting (MV) and weighted average (WA). The experimental results are evaluated with well-known databases, the '11k Hands' and the Hong Kong Polytechnic University Contactless Hand Dorsal Images 'PolyU', show the proposed algorithm's efficiency. Using the MV on the Bray-Curtis similarity measure, the fingernail-based and the base-knuckle- based fusion obtained 100% in the identification estimation. In addition, the identification rate gained 100% in regions of hands and whole hands from the two popular datasets exceeded the performance of the state-of-the-art approaches.

Item Type:
Contribution in Book/Report/Proceedings
Additional Information:
©2022 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
ID Code:
187472
Deposited By:
Deposited On:
08 Mar 2023 13:30
Refereed?:
Yes
Published?:
Published
Last Modified:
23 Oct 2024 23:28