P300 Classification with ConvNets for Brain Invader

Brain–computer interfaces (BCI) record brain activity with sensors like electroencephalography (EEG) to turn into commands to machines. There are particular signals like the P300 that appear a few milliseconds after witnessing a surprising event like a flash on a screen. Researchers have developed procedures to implement BCI systems in entertainment applications, such as video games, to enhance user experience and provide opportunities for disabled persons. Deep learning is one of the leading methods to decode EEG signals with convolutional neural networks. In spirit, we propose a new architecture based on famous architectures such as EEGNet, ResNet, and ResNext for classification. We propose to use the Brain Invader dataset to test our hypothesis. The results proved that our proposition outperforms state-of-the-art methods like Riemannian Geometry, EEGNet, and others.