Article Published in International Conference on Neural Information Processing (ICONIP 2018)

November, 2018 SDU published the work “Development of a Real-Time Motor-Imagery-Based EEG Brain-Machine Interface

Abstract

EEG-based brain-machine interfaces offer an alternative means of interaction with the environment relying solely on interpreting brain activity. They can not only significantly improve the life quality of people with neuromuscular disabilities, but also present a wide range of opportunities for industrial and commercial applications. This work focuses on the development of a real-time brain-machine interface based on processing and classification of motor imagery EEG signals. The goal was to develop a fast and reliable system that can function in everyday noisy environments. To achieve this, various filtering, feature extraction, and classification methods were tested on three data sets, two of which were recorded in a noisy public setting. Results suggested that the tested linear classifier, paired with band power features, offers higher robustness and similar prediction accuracy, compared to a non-linear classifier based on recurrent neural networks. The final configuration was also successfully tested on a real-time system.

Link to download