Investigation of a Deep-Learning Based Brain–Computer Interface With Respect to a Continuous Control Application

As part of a motor-imagery brain-computer interface (BCI), deep neural network (DNN) must analyze measured electroencephalogram (EEG) data and identify signal patterns characteristic particular imagined motor movement. Our studies are intended to investigate the use such DNN in asynchronous online applications, where EEG signals need be interpreted continuously, as well gain insights into learned patterns. We examined EEGNet, commonly referenced convolutional net (CNN). In addition impacts size temporal position trials used for training testing on classification accuracy, we contributions behavior known their effects response time system period which mental state was stably recognized. Because optimal is different involved, introduced ‘cropped training’, method trained using with positions. This enabled learn 0–8 Hz frequency range that important short 8–30 determining duration. show cropped essential achieving good