Evaluation of Consumer-Grade EEG Headsets for BCI Drone Control

Abstract

Brain-computer interface (BCI) allows people to control a computer system using their brain signals. The recent availability of consumer-grade electroencephalography (EEG) headsets enables this technology to be used outside the lab. In particular, there has been interest in controlling a drone using brain signals. We propose a system that allows users to manipulate the thrust of the drone using consumer-grade EEG headsets. Active concentration, represented by EEG band powers, was investigated as an input modality for the BCI system. Comparisons were also made between different approaches in data retrieval and processing, as well as headsets. It was found that using supervised learning with SVM and the Muse headset resulted in better performance (around 70%). Offline evaluation shows that while both Emotiv Insight and Muse had comparable accuracy, Muse had better usability and was thus adopted in our system. Online user testing with the implemented BCI system revealed variable performance across subjects. This highlights the need for incremental training of both classifier and users for improved efficacy.