Scholarlink

To address the challenges of feature instability and domain shift in palmprint key generation under cross-dataset and open-set scenarios, this paper proposes the PalmOKey framework. The framework extracts discriminative and domain-invariant features through a Single-Source Domain Generalization (SSDG)-enhanced Comprehensive Competition Network (CCNet) and combines a fuzzy commitment scheme based on Neural Low-Density Parity-Check (LDPC) codes to achieve reliable key generation. Experimental results show that PalmOKey performs excellently in cross-dataset key generation tasks: when MS_Blue is used as the source dataset and Tongji as the target dataset, the True Acceptance Rate (TAR) at a False Acceptance Rate (FAR) of 0% reaches 99.97%; the Equal Error Rate (EER) on the Tongji dataset is as low as 0.0000, significantly outperforming existing methods.

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

Cross-Domain Palmprint Cryptosystems via Image Alignment and Neural Error Correction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yanhong Qian,
  • Xinyue Liu,
  • Fuyou Leng,
  • Hui Zhang,
  • Xingbo Dong,
  • Zhe Jin

摘要

To address the challenges of feature instability and domain shift in palmprint key generation under cross-dataset and open-set scenarios, this paper proposes the PalmOKey framework. The framework extracts discriminative and domain-invariant features through a Single-Source Domain Generalization (SSDG)-enhanced Comprehensive Competition Network (CCNet) and combines a fuzzy commitment scheme based on Neural Low-Density Parity-Check (LDPC) codes to achieve reliable key generation. Experimental results show that PalmOKey performs excellently in cross-dataset key generation tasks: when MS_Blue is used as the source dataset and Tongji as the target dataset, the True Acceptance Rate (TAR) at a False Acceptance Rate (FAR) of 0% reaches 99.97%; the Equal Error Rate (EER) on the Tongji dataset is as low as 0.0000, significantly outperforming existing methods.