Finger vein recognition based on efficient channel attention and GhostNet
by Yintao Ke; Hui Zheng; Jing Jie; Beiping Hou; Yuchuan Chen
International Journal of Wireless and Mobile Computing (IJWMC), Vol. 26, No. 2, 2024

Abstract: The deep convolutional network (DCN) suffers from drawbacks such as high computational complexity and slow speed. To address these issues and facilitate the deployment of DCNs in embedded devices, we propose a finger vein recognition method based on lightweight Efficient Channel Attention (ECA) mechanism and GhostNet. By combining the ECA mechanism with the GhostNet's G-bneck, we create a new module called ECAGhostNet. Additionally, we establish a more realistic FV-UST dataset for finger vein door locks which includes images with various challenges like rotation, stains, skin damage, hand sweat, temperature variations, and illumination differences. Experimental results demonstrate that ECAGhostNet outperforms GhostNet on the public FV-USM dataset, improving accuracy by 0.82% with minimal parameter increase (1.9 M). On the self-built FV-UST dataset, ECAGhostNet achieves a 0.63% accuracy improvement over GhostNet. Furthermore, we validate the effectiveness of our proposed model on the Jetson Nano device, confirming its suitability for real-world embedded applications.

Online publication date: Mon, 04-Mar-2024

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Wireless and Mobile Computing (IJWMC):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com