Scholarlink

Biometric authentication systems are essential for secure access in various domains and are increasingly integrated into everyday life. By implementing the multimodal technology, we enhance the security by combining multiple traits like fingerprints and facial recognition, addressing the limitations of unimodal systems, which are prone to spoofing and errors. This paper presents a multimodal biometric authentication system that combines finger vein and finger knuckle modalities. Preprocessing involves Bilateral Filtering followed by Adaptive Histogram Equalization (AHE) optimized with the Osprey Optimization Algorithm (OOA). Feature selection leverages an improved novel adaptive SHO and a novel hybrid of adaptive Sea-Horse Optimization with the Walrus Optimization Algorithm (aSHO-WaOA). Finger-vein images are classified using ShuffleNet, and finger-knuckle images are processed with MobileNetV3. Finally, score-level fusion integrates the classification results for enhanced accuracy and reliability. The performance of the system is evaluated on accuracy. The databases used to train the model are IIT Delhi Finger Knuckle Dataset and Hong Kong Polytechnic Finger Image Dataset. The proposed hybrid aSHO-WaOA algorithm outperforms other nature inspired algorithms in terms of precision and convergence, thus providing a robust solution for multimodal biometric recognition making it secure and reliable.

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

Transfer Learning-Based Multimodal Biometric Recognition System Using Osprey Optimized AHE and a Novel Hybrid of Adaptive SHO and WaOA for Feature Selection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Indu Singh,
  • Sawan Jain,
  • Kartikey Saxena,
  • Sarvesh Pahwa,
  • Shaurya Wadehra

摘要

Biometric authentication systems are essential for secure access in various domains and are increasingly integrated into everyday life. By implementing the multimodal technology, we enhance the security by combining multiple traits like fingerprints and facial recognition, addressing the limitations of unimodal systems, which are prone to spoofing and errors. This paper presents a multimodal biometric authentication system that combines finger vein and finger knuckle modalities. Preprocessing involves Bilateral Filtering followed by Adaptive Histogram Equalization (AHE) optimized with the Osprey Optimization Algorithm (OOA). Feature selection leverages an improved novel adaptive SHO and a novel hybrid of adaptive Sea-Horse Optimization with the Walrus Optimization Algorithm (aSHO-WaOA). Finger-vein images are classified using ShuffleNet, and finger-knuckle images are processed with MobileNetV3. Finally, score-level fusion integrates the classification results for enhanced accuracy and reliability. The performance of the system is evaluated on accuracy. The databases used to train the model are IIT Delhi Finger Knuckle Dataset and Hong Kong Polytechnic Finger Image Dataset. The proposed hybrid aSHO-WaOA algorithm outperforms other nature inspired algorithms in terms of precision and convergence, thus providing a robust solution for multimodal biometric recognition making it secure and reliable.