View in EDS

Flexible joint sparse representation: enhancing multimodal biometric verification accuracy and anti-spoofing.

Saved in:
Bibliographic Details
Title: Flexible joint sparse representation: enhancing multimodal biometric verification accuracy and anti-spoofing.
Authors: Wang, Yi1 (AUTHOR) luckywangyi@stu.xhu.edu.cn, Chen, Dongliang1 (AUTHOR) 212024081200023@stu.xhu.edu.cn, Huang, Zengxi1,2 (AUTHOR) huangzx001@xhu.edu.cn, Jia, Nian1 (AUTHOR) jianian@163.com, Song, Fei1 (AUTHOR) sfei_work@mail.xhu.edu.cn, Ma, Tingsong1 (AUTHOR) mts@mail.xhu.edu.cn, Zhu, Changyu1,2 (AUTHOR) icic_icic@163.com
Source: Multimedia Systems. Jun2026, Vol. 32 Issue 4, p1-15. 15p.
Subjects: Biometric identification, Sparse approximations, Optimization algorithms, Mathematical regularization, Classification algorithms
Abstract: In the field of biometric recognition, enhancing the accuracy and anti-spoofing performance of multimodal verification systems remains a critical challenge. Previous research has demonstrated the considerable potential of sparse representation-based classification (SRC) in multimodal verification by incorporating non-target subjects and performing one-to-many competitive matching. However, existing joint sparse representation models either impose overly strict constraints on atom-level sparsity patterns across modalities or grant excessive freedom in sparse coding. To address this issue, this paper proposes a Flexible Joint Sparse Representation Model (FMJSR) designed for multimodal systems utilizing different biometric traits. FMJSR promotes class-level sparsity consistency while allowing atom-level flexibility. This design is motivated by the observation that, for a pair of multimodal query images (e.g., face and palmprint), the order of their best-matching gallery templates within individual sub-dictionaries often differs, yet their class labels remain consistent for a genuine user. The model employs a mixed-norm regularization to enforce class-level consistency and an atom-level norm to accommodate modality-specific variations. Optimization is carried out using the Alternating Direction Method of Multipliers (ADMM). Moreover, we introduce a novel dual-examination decision mechanism that combines the sparsity-based matching score and its ranking information for final verification to promote security. Experimental results on multimodal datasets containing face, ear, and palmprint data show that FMJSR outperforms existing methods in both verification accuracy and anti-spoofing capability, achieving nearly 0% Equal Error Rate (EER) in licit scenarios and significantly lower Spoof EER (SEER) under partial spoofing attacks.The code is available at:https://github.com/wangyidaoyyds/FMJSR [ABSTRACT FROM AUTHOR]
Copyright of Multimedia Systems is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:In the field of biometric recognition, enhancing the accuracy and anti-spoofing performance of multimodal verification systems remains a critical challenge. Previous research has demonstrated the considerable potential of sparse representation-based classification (SRC) in multimodal verification by incorporating non-target subjects and performing one-to-many competitive matching. However, existing joint sparse representation models either impose overly strict constraints on atom-level sparsity patterns across modalities or grant excessive freedom in sparse coding. To address this issue, this paper proposes a Flexible Joint Sparse Representation Model (FMJSR) designed for multimodal systems utilizing different biometric traits. FMJSR promotes class-level sparsity consistency while allowing atom-level flexibility. This design is motivated by the observation that, for a pair of multimodal query images (e.g., face and palmprint), the order of their best-matching gallery templates within individual sub-dictionaries often differs, yet their class labels remain consistent for a genuine user. The model employs a mixed-norm regularization to enforce class-level consistency and an atom-level norm to accommodate modality-specific variations. Optimization is carried out using the Alternating Direction Method of Multipliers (ADMM). Moreover, we introduce a novel dual-examination decision mechanism that combines the sparsity-based matching score and its ranking information for final verification to promote security. Experimental results on multimodal datasets containing face, ear, and palmprint data show that FMJSR outperforms existing methods in both verification accuracy and anti-spoofing capability, achieving nearly 0% Equal Error Rate (EER) in licit scenarios and significantly lower Spoof EER (SEER) under partial spoofing attacks.The code is available at:https://github.com/wangyidaoyyds/FMJSR [ABSTRACT FROM AUTHOR]
ISSN:09424962
DOI:10.1007/s00530-026-02214-z