An Auditory Brain Computer Interface System Based on Naturalistic Paradigm

Naturalistic paradigms, which incorporate sensory stimuli commonly encountered in daily life, have garnered widespread attention in psychology and neuroscience in recent years. This study proposed an auditory brain-computer interface (BCI) system based on naturalistic paradigm. The system uses the reading of the article as a background while employing the sounds of chicken and dog as stimuli to evoke ERP from the subjects. To further enhance the performance of the BCI system, we improved the EEGNet model, which is based on convolutional neural networks, by incorporating a prototypical classifier to improve the classification performance in small-sample scenarios. Experimental results demonstrate that the proposed paradigm with background evokes P300 components with higher amplitudes compared to the paradigm without background, thus improving the classification performance of BCI. Additionally, the modifications made to EEGNet further improved the model’s classification performance. Our proposed brain-computer interface system achieved an average online accuracy of 82.4% in eight participants, with five out of the eight participants achieving an accuracy of over 85%. The auditory BCI system proposed in this study may potentially assist patients with disorders of consciousness (DoC) in the future.