Preliminary Study of EEG-based Brain–Computer Interface Systems for Assisted Mobility Applications

As a consequence of an increasingly elderly population, accommodating age-related impairments has become a serious concern in many countries. To assist people with disabilities, the use of electroencephalography (EEG)-based brain–computer interface (BCI) systems has been explored and has proven to be both clinically usable and feasible. In this study, four mental tasks that are directly related to mobility commands were implemented into the simulated design of a BCI-controlled exoskeleton. These mental tasks were used in a BCI training procedure. EEG signals were recorded using a modified emotive headset and were analysed in the time–frequency domain. CSP patterns were used to evaluate the data acquired during the completion of the mental tasks. A classifier that is based on the Support Vector Machines (SVM) method was used. The average classification error for the ten participants was 0.411 and the lowest error rate was 0.304. In the two–two contrast groups, two different patterns from one group showed a large area of overlap, indicating poor performance of the CSP method for this BCI training procedure. This may have been caused by the motor imagery area overlap of the four mental tasks. Practitioner Summary: The high average classification error implies that while the chosen complex mental tasks may be appropriately chosen for BCI training procedure, the features of comparison should be chosen more carefully. More research into the features of these mental tasks is required as they have a large influence on their classification results.