Going Deeper with Recurrent Convolutional Neural Networks for ClassifyingP300 BCI Signals

We develop and test three deep-learning recurrent convolutional architectures forlearning to recognize single trial EEG event-related potentials for P300 brain-computerinterfaces (BCI)s. The existing classifiers for P300 detection don't preserve the spatiotemporalstructure of the data, thereby losing local spatial and temporal patterns in thedata. The proposed models respect the spatial and temporal nature of the EEG signals.A three-dimensional convolutional neural network (3D-CNN) is used in concertwith a two-dimensional convolutional neural network (2D-CNN) and LSTM to capturethe spatiotemporal patterns in the EEG signals. Moreover, a transfer learning based approach is applied while training the subjects. One advantage of the neural networksolution is that it provides a natural way to share a lower-level feature space betweensubjects while adapting the classifier that works on that feature space. We compare thedeep neural networks with the standard methods for P300 BCI classification.