Abstract
Deciphering electroencephalogram (EEG) signals accurately poses a formidable challenge due to their intrinsic high dimensionality, non-stationarity, and intricate spatiotemporal patterns. While convolutional neural networks (CNNs) have found widespread use in EEG signal processing, their limited receptive fields impede their capacity to capture long-range dependencies, which are pivotal for comprehensive EEG analysis. To overcome this constraint, this paper introduces a novel hybrid intelligent brainwave recognition model that amalgamates convolutional layers with a transformer-based self-attention mechanism for EEG signal interpretation. The proposed model harnesses the strengths of convolutional layers to grasp local spatiotemporal features, while utilizing self-attention to effectively discern global correlations in EEG signals. Evaluation The efficacy of the proposed approach was assessed on the Physionet EEG dataset, yielding an accuracy of 88.7% and a Kappa score of 86.3%, surpassing existing methods solely reliant on CNNs. These findings underscore the promise of hybrid architectures in robust EEG signal recognition and their potential utility in clinical settings and brain-computer interface applications.