Advancing EEG Classification with Transformer Architecture: A Case Study on P300 Potentials for Assistive Devices

In this investigation, the Transformer architecture is applied to classify P300 event-related potentials in publicly available EEG data to augment assistive communication devices for individuals with impaired communication abilities. A detailed analysis comparing the Transformer model to established algorithms documented in the literature demonstrates its effectiveness in terms of classification accuracy, achieving a notable 95% accuracy rate, indicating its viability as a classifier for P300-based spellers. This result underlines the potential of advanced AI techniques in enhancing neurotechnology applications and suggests a new benchmark in the field. Future work will focus on refining these methods to further improve the usability and performance of assistive devices, aiming to bridge the gap between AI advancements and practical healthcare applications. This endeavour contributes to the biomedical engineering and artificial intelligence disciplines by offering insights into EEG data analysis and its implications for developing more accessible assistive technologies.