Electroencephalography data analysis with convolutional and recurrent neural networks

Modern methods for the neurophysiological data analysis provide promising solutions to various problems both in the field of medical industry and in the field of brain–computer interfaces development. In this paper, a couple of important electroencephalography (EEG) data analysis problems are considered that are artifact detection and removal and human emotion recognition. Due to recent active development of algorithms based on deep artificial neural networks and to the cost reduction of commercial prototypes of brain–computer interfaces, the efficiency and robustness of modern methods for EEG data analysis is approaching a level sufficient for use outside the laboratory. The paper proposes methods for EEG data analysis based on convolutional and recurrent neural networks which make it possible to achieve high accuracy of artifact classification and emotion recognition over open data sets.