Optimization of Electroencephalogram Feature Based on Multi-Band Feature Matrix and Relative Difference for Imagined Motor Movement Pattern Classification

Abstract

Electroencephalogram (EEG)-based imagined motor movement recognition plays a crucial role in brain–computer interface applications and medical rehabilitation systems. However, EEG signals inherently exhibit high levels of noise, are nonlinear and nonstationary, contain spatial information, and show significant signal pattern variance across individuals, complicating feature extraction and classification. This study proposes a novel framework that integrates the relative difference method with a multi-band feature matrix, explicitly incorporating four EEG frequency bands (theta, alpha, beta, and gamma) mapped according to the international system 10–20. The extracted features are processed by a three-layer convolutional neural network (CNN) for classification. The proposed method is evaluated on the MIMED dataset for six-class imagined motor movement recognition and validated on the AMIGOS dataset for four-class emotion recognition. The experimental evaluation on the MIMED dataset demonstrates consistently strong classification performance, achieving 99.36% accuracy, 99.37% precision, 99.37% recall, and an F1-score of 99.36%. These results significantly outperform those without baseline reduction, the difference, and the fractional methods. Further validation on AMIGOS shows that the proposed model achieves an accuracy of 99.64%. This result surpasses the previously reported state of the art. These findings confirm that combining relative difference and multi-band feature matrices produces more stable, representative, and discriminative EEG features, making them readily recognizable by CNN methods. The proposed approach has strong potential for real-time brain–computer interfaces, clinical neurorehabilitation, and emotion-sensitive human–computer interaction systems



Received: 10 August 2025 | Revised: 14 February 2026 | Accepted: 10 April 2026



Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.



Data Availability Statement

The MIMED dataset that supports the findings of this study is openly available in Mendeley Data at https://data.mendeley.com/datasets/zs25xxjkm9/3. The AMIGOS dataset that supports the findings of this study is openly available at https://doi.org/10.1109/TAFFC.2018.2884461, reference number [18].



Author Contribution Statement

I Made Agus Wirawan: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. I Nyoman Sukajaya: Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Ni Nyoman Mestri Agustini: Methodology, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration.