TongueBCI: An Interaction Method Based on EEG Signals from Tongue Movement Direction

Computer control based on tongue gestures has been shown to be a potentially accessible interaction method for patients with paralyzing injuries. Considering the advantages of nonintrusive brain-computer interfaces (BCIs), this study explored novel tongue-computer interaction methods by analyzing the electroencephalogram (EEG) signals of tongue gestures. We developed TongueBCI, a nonintrusive tongue-computer interface that could recognize a user’s interaction intention using EEG signals from two types of tongue gestures-that is, Motor Imagery (MI) and Motor Execution (ME). We first constructed an EEG dataset that included both types of tongue gestures and then developed a deep-learning model, which achieved an accuracy of 94.1%. We then implemented two corresponding real-time interaction technologies based on MI/ME tongue gestures and designed an experimental system to test their effectiveness. The results showed that MI is a more suitable approach for facilitating accessible interactions than ME.