The Relationship Between the Movement Difficulty and Brain Activity Before Arm Movements

摘要

A brain-computer interface (BCI) is a technology that can control external devices using brain activity. It is expected that the flexibility and safety of a BCI will be improved if movement-related information can be extracted from brain activity before executing the movement. In this study, we examined whether movement difficulty levels can be decoded from electroencephalogram (EEG) data. We conducted an experiment where in five participants performed arm reaching movements with three different levels of difficulty, brain activity was measured before these movements. To classify the levels of difficulty, we extracted event-related spectrum perturbation (ERSP) data and performed classification using a relevance vector machine (RVM). Single-trial classification using ERSP data could not obtain high classification accuracy. However, classification accuracies using averaged-trial ERSP data were 66.0% on average (53.9%, 82.3%, 79.6%, 53.1% and 61.1% for each participant). These results show that information related to movement difficulty might be decoded from brain activity before movement, although it is necessary to improve the performance at the single-trial level in future work.