No abstract

Many people suffer from movement disabilities and would benefit from an assistive mobility device with practical control. This paper demonstrates a face-machine interface system that uses motion artifacts from electroencephalogram (EEG) signals for mobility enhancement in people with quadriplegia. We employed an Emotiv EPOC X neuroheadset to acquire EEG signals. With the proposed system, we verified the preprocessing approach, feature extraction algorithms, and control modalities. Incorporating eye winks and jaw movements, an average accuracy of 96.9% across four commands was achieved. Moreover, the online control results of a simulated power wheelchair showed high efficiency based on the time condition. The combination of winking and jaw chewing results in a steering time on the same order of magnitude as that of joystickbased control, but still about twice as long. We will further improve the efficiency and implement the proposed face-machine interface system for a real-power wheelchair.

The research on the electroencephalography (EEG)-based brain–computer interface (BCI) is widely utilized for wheelchair control. The ability of the user is one factor of BCI efficiency. Therefore, we focused on BCI tasks and protocols to yield high efficiency from the robust EEG features of individual users. This study proposes a task-based brain activity to gain the power of the alpha band, which included eyes closed for alpha response at the occipital area, attention to an upward arrow for alpha response at the frontal area, and an imagined left/right motor for alpha event-related desynchronization at the left/right motor cortex. An EPOC X neuroheadset was used to acquire the EEG signals. We also proposed user proficiency in motor imagery sessions with limb movement paradigms by recommending motor imagination tasks. Using the proposed system, we verified the feature extraction algorithms and command translation. Twelve volunteers participated in the experiment, and the conventional paradigm of motor imagery was used to compare the efficiencies. With utilized user proficiency in motor imagery, an average accuracy of 83.7% across the left and right commands was achieved. The recommended MI paradigm via user proficiency achieved an approximately 4% higher accuracy than the conventional MI paradigm. Moreover, the real-time control results of a simulated wheelchair revealed a high efficiency based on the time condition. The time results for the same task as the joystick-based control were still approximately three times longer. We suggest that user proficiency be used to recommend an individual MI paradigm for beginners. Furthermore, the proposed BCI system can be used for electric wheelchair control by people with severe disabilities.

No abstract