E-SAT: An extreme learning machine based self attention approach for decoding motor imagery EEG in subject-specific tasks

Journal of Neural Engineering, Journal Year: 2024, Volume and Issue: 21(5), P. 056033 - 056033

Published: Oct. 1, 2024

The advancements in Brain-Computer Interface (BCI) have substantially evolved people's lives by enabling direct communication between the human brain and external peripheral devices. In recent years, integration of machine larning (ML) deep learning (DL) models considerably imrpoved performances BCIs for decoding motor imagery (MI) tasks. However, there still exist several limitations, e.g., extensive training time high sensitivity to noises or outliers with those existing models, which largely hinder rapid developments BCIs. To address such issues, this paper proposes a novel extreme (ELM) based self-attention (E-SAT) mechanism enhance subject-specific classification performances. Specifically, E-SAT, ELM is employed both imrpove module generalization ability feature extraction optimize model's parameter initialization process. Meanwhile, extracted features are also classified using ELM, end-to-end setup used evaluate E-SAT on different MI EEG signals. Extensive experiments datasets, as BCI Competition III Dataset IV-a, IV-b IV Datasets 1,2a,2b,3, conducted verify effectiveness proposed strategy. Results show that outperforms state-of-the-art (SOTA) methods all an average accuracy 99.8%,99.1%,98.9%,75.8%, 90.8%, 95.4%, being achieved each respectively. experimental results not only outstanding performance extractions, but demonstrate it helps achieves best among nine other robust ones. addition, study exceptional binary multi-class tasks, well noisy non-noisy datatsets. .

Language: Английский

Evolving Trends and Future Prospects of Transformer Models in EEG-Based Motor-Imagery BCI Systems

Published: Jan. 1, 2024

Language: Английский

A Dual Adaptation Approach for EEG-Based Biometric Authentication Using the Ensemble of Riemannian Geometry and NSGA-II

Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 91 - 109

Published: Dec. 1, 2024

Language: Английский