A High-Performance Classification Method Based on CNN for Interventional EEG Data

Interventional brain-computer interfaces (BCI) have been developed for many years and are capable of collecting electroencephalogram (EEG) signals. However, research on interventional EEG signals is just beginning. In this paper. We present a new deep learning network for intentional classification of interventional EEG in a monkey. Electrodes are placed on the inner walls of the vessels in the superior sagittal sinus to collect signals in four states. Then, we construct a Convolutional Neural Networks (CNN) model to extract features and classify the states. In the experiments, we recognize the interventional EEG signals of the monkey in the four states, and the average recognition accuracy of its EEG signals in each state is more than 90% for the whole dataset, and 70% for the validation set. The experimental results show that the method in this paper can effectively recognize and classify interventional EEG signals in different states. Our work advances the application of interventional BCIs and paves the way for the future development of more versatile and responsive interventional EEG interfaces.