Multichannel EEG-Based Motor Movement Classification Using Fusion of 1D-CNN and Attention Mechanism

Brain Computer Interface (BCI) technologies use Electroencephalogram (EEG) impulses to provide direct interaction between the brain and external equipment, therefore presenting a revolutionary opportunity for persons with movement disabilities. Despite considerable progress, motor imagery classification has encountered substantial obstacles, such as a strong sensitivity to noise, non-stationarity in EEG signals, and the need for comprehensive human feature engineering. These limits sometimes reduce precision and applicability across individuals and sessions. To address these and improve the efficiency and efficacy of categorization, this study introduces an integrated deep learning framework that fuses 1D Convolutional Neural Networks (CNN) with a soft attention module. The objective of this framework is to autonomously extract and rank pertinent characteristics from multichannel EEG inputs. The system especially targets the classification of right and left fist movements using the EEG Motor Movement/Imagery dataset. The EEG data is refined using preprocessing procedures, including Min-Max normalization and segmentation, to enhance its suitability for feature extraction. The proposed model utilizes CNNs to effectively capture spatial patterns in the EEG data. Additionally, it incorporates the attention mechanism to highlight important characteristics, leading to improved classification performance. By attaining an overall test accuracy of 82%, this model surpasses conventional methods by minimizing manual feature engineering and boosting generalization across subjects and sessions, therefore showcasing its potential for practical use in BCI systems and propelling the field toward more dependable and efficient EEG-based interfaces.