| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 4344336 | Neuroscience Letters | 2012 | 6 Pages |
Most brain–computer interfaces (BCIs) are non-time-restraint systems. However, the method used to design a real-time BCI paradigm for controlling unstable devices is still a challenging problem. This paper presents a real-time feedback BCI paradigm for controlling an inverted pendulum on a cart (IPC). In this paradigm, sensorimotor rhythms (SMRs) were recorded using 15 active electrodes placed on the surface of the subject's scalp. Subsequently, common spatial pattern (CSP) was used as the basic filter to extract spatial patterns. Finally, linear discriminant analysis (LDA) was used to translate the patterns into control commands that could stabilize the simulated inverted pendulum. Offline trainings were employed to teach the subjects to execute corresponding mental tasks, such as left/right hand motor imagery. Five subjects could successfully balance the online inverted pendulum for more than 35 s. The results demonstrated that BCIs are able to control nonlinear unstable devices. Furthermore, the demonstration and extension of real-time continuous control might be useful for the real-life application and generalization of BCI.
Graphical abstractFigure optionsDownload full-size imageDownload high-quality image (174 K)Download as PowerPoint slideHighlights► Five subjects successfully balanced the inverted pendulum by motor imagery. ► Online adaptation and subject training were necessary in SMR-BCIs. ► Subject's control strategy also played important role in stabilizing the pendulum. ► Fatal lapse was usually made by the subject rather than the system. ► BCIs can be employed to control unstable devices such as an inverted pendulum.
