Advancing EEG-based biometric identification through multi-modal data fusion and deep learning techniques

• Published in: Complex & intelligent systems Vol. 11; no. 9; pp. 398 - 19
• Main Authors: Rehman, Touseef Ur, Alruwaili, Madallah, Siddiqi, Muhammad Hameed, Alhwaiti, Yousef, Anwar, Sajid, Halim, Zahid, Alam, Maaz
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.09.2025; Springer Nature B.V; Springer
• Subjects: Access control; Artificial intelligence; Biometric identification; Biometrics; Brain research; Complex systems; Complex systems optimization; Complexity; Computational Intelligence; Cybersecurity; Data analysis; Data integration; Data Structures and Information Theory; Data transmission; Deep learning; EEG-based biometric identification; Electroencephalography; Engineering; Identification systems; Investigations; Machine learning; Modal data; Multi-modal data fusion; Neural networks; Neurosciences; Original Article; Pattern recognition; Physiology; Self organizing systems; Deep learning; EEG-based biometric identification; Pattern recognition; Complex systems optimization; Multi-modal data fusion
• DOI: 10.1007/s40747-025-02012-6
• ISSN: 2199-4536

The integration of diverse data modalities is critical for advancing the understanding and optimization of complex systems. In this context, EEG-based biometric identification represents a unique challenge and opportunity for multi-modal data fusion. EEG signals, characterized by their high complexity and variability, offer a non-intrusive and reliable means of individual identification. This work proposes an advanced deep learning-based framework to extract and analyze distinctive EEG frequency patterns, enhancing the accuracy and robustness of EEG-based biometric systems. Two experimental setups were designed to evaluate the intelligent fusion of EEG data across varied brain activity tasks. In the first setup, the model was trained on data from subjects performing a single task, then assessed on its generalization across diverse tasks, demonstrating its ability to adapt to heterogeneous data streams. This methodology achieved a biometric recognition accuracy of up to 99%, highlighting the potential of intelligent data integration techniques in uncovering hidden patterns within complex physiological data. By leveraging the synergy of multi-modal data analysis and deep learning, this work contributes to the broader objective of developing self-organizing systems capable of adapting to diverse data sources. These findings underscore the transformative potential of EEG-based biometrics within the broader domain of multi-modal data fusion, offering promising applications in healthcare, security, and beyond.