Article: Attention-enhanced linear discriminant analysis for robust facial recognition in dynamic attendance systems Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.19 pp.16 - 37 Abstract: In the dynamic attendance scenario, the human face images suffer from illumination variations and posture differences, which leads to the decline of the discriminative power of the traditional linear discriminant analysis method. This paper proposes a linear discriminant analysis algorithm embedded with an attention module. This method uses the attention mechanism to adaptively focus on the key feature regions of the human face, effectively suppressing irrelevant background interference, thereby enhancing the robustness of the model in complex environments. Experimental results on the public datasets labelled faces in the wild and CelebFaces Attributes Dataset show that the recognition accuracy of this method reaches 98.7% and 96.2% respectively, which is 3.5% and 4.1% higher than the classic linear discriminant analysis method. This research provides an effective solution for improving the reliability of identity verification in non-cooperative scenarios. Inderscience Publishers - linking academia, business and industry through research

Title: Attention-enhanced linear discriminant analysis for robust facial recognition in dynamic attendance systems

Authors: Yazhong Cui; Ming Yan; Chaoyi Zhou

Addresses: Shendong Coal Group Co., Ltd., Yulin, 719315, China ' Shaanxi Yijiexin Information Technology Co., Ltd., Xi'an, 710065, China ' Shendong Intelligent Technology Service, Yulin, 719315, China

Abstract: In the dynamic attendance scenario, the human face images suffer from illumination variations and posture differences, which leads to the decline of the discriminative power of the traditional linear discriminant analysis method. This paper proposes a linear discriminant analysis algorithm embedded with an attention module. This method uses the attention mechanism to adaptively focus on the key feature regions of the human face, effectively suppressing irrelevant background interference, thereby enhancing the robustness of the model in complex environments. Experimental results on the public datasets labelled faces in the wild and CelebFaces Attributes Dataset show that the recognition accuracy of this method reaches 98.7% and 96.2% respectively, which is 3.5% and 4.1% higher than the classic linear discriminant analysis method. This research provides an effective solution for improving the reliability of identity verification in non-cooperative scenarios.

Keywords: dynamic attendance; facial recognition; linear discriminant analysis; LDA; attention module; feature discriminative power.

DOI: 10.1504/IJRIS.2026.155642

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.19, pp.16 - 37

Received: 10 Feb 2026
Accepted: 14 Mar 2026
Published online: 07 Aug 2026 *

Title: Attention-enhanced linear discriminant analysis for robust facial recognition in dynamic attendance systems

Keep up-to-date