Single-Channel EEG Classification for Motor Tasks via Bayesian-Optimized Lightweight 1D CNN

Electroencephalogram (EEG) monitoring enables the capture of brain activity and provides insights into motor tasks with applications in rehabilitation, prosthetic control, and Brain-Computer Interfaces (BCI). Traditional classification approaches often rely on complex architectures, limiting usability in resource-constrained settings. This study introduces a 1-dimensional Convolutional Neural Network (1D CNN) optimised using Gaussian-based Bayesian optimisation for classifying single-channel EEG signals. While existing models are deep, requiring extensive pre-and post-processing, the proposed model achieves a trade-off between computational efficiency and classification accuracy. Trained exclusively on a private dataset of 27 individuals performing motor tasks, the model achieved a promising accuracy of 81.82%, demonstrating its potential for practical deployment. The proposed model also outperformed complex pre-trained architectures such as VGG-16 and GoogLeNet by achieving significantly fewer parameters and lower computational demand, making it particularly suitable for resource-limited environments. As an early investigation, this work explores the feasibility of single-channel approaches using lightweight architectures for real-time EEG classification in resource-constrained environments.