Scholarlink

<p>Finger vein recognition has emerged as a promising biometric technique due to its high security, individuality, and contactless nature. Despite progress, challenges remain in effectively extracting and aggregating both global and local finger vein features. To address these, we propose a bi-branch multi-level feature aggregation network (BMFAN). BMFAN enhances recognition accuracy by extracting multi-level features, refining local and global features through dedicated units, and aggregating them using an attention mechanism. Experimental results on the SDUMLA and HKPU datasets demonstrate the effectiveness of our approach, achieving recognition accuracies (ACCs) of 99.37% and 99.35%, respectively, with equal error rates (EERs) of 0.21% and 0.08%. These findings highlight the importance of aggregating global and local features for intricate finger vein recognition. Codes and datasets are available at <a href="https://github.com/lipeifeng1101/BMFAN.git.">https://github.com/lipeifeng1101/BMFAN.git.</a></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced finger vein recognition via Bi-branch multi-level feature aggregation network

  • 期刊论文

  • 发表日期：

    2025年12月11日

查看全文
  • Peifeng Li,
  • Xianjing Meng,
  • Hengwu Li,
  • Chi Zhang,
  • Changhao Dou

摘要

Finger vein recognition has emerged as a promising biometric technique due to its high security, individuality, and contactless nature. Despite progress, challenges remain in effectively extracting and aggregating both global and local finger vein features. To address these, we propose a bi-branch multi-level feature aggregation network (BMFAN). BMFAN enhances recognition accuracy by extracting multi-level features, refining local and global features through dedicated units, and aggregating them using an attention mechanism. Experimental results on the SDUMLA and HKPU datasets demonstrate the effectiveness of our approach, achieving recognition accuracies (ACCs) of 99.37% and 99.35%, respectively, with equal error rates (EERs) of 0.21% and 0.08%. These findings highlight the importance of aggregating global and local features for intricate finger vein recognition. Codes and datasets are available at https://github.com/lipeifeng1101/BMFAN.git.