A brain-computer interface system for lower-limb exoskeletons based on motor imagery and stacked ensemble approach

Area of Science:

• Neuroscience
• Robotics
• Rehabilitation Engineering

Background:

• Existing lower limb exoskeletons (LLEs) often lack sufficient patient engagement during rehabilitation.
• Integrating patient motion intentions into LLE control is crucial for effective therapy.

Purpose of the Study:

• To develop an online brain-computer interface (BCI) system for LLEs that enhances patient involvement.
• To improve the precision of LLE control through advanced signal decoding techniques.

Main Methods:

• An online BCI system was established for closed-loop control, including brain signal collection, decoding, robotic control, and real-time feedback.
• A novel classification algorithm, weighted random forests-support vector machines (WRF-SVM), was proposed using stacked ensemble learning.
• An online experimental protocol incorporating visual and proprioceptive feedback was designed.

Main Results:

• The proposed WRF-SVM algorithm demonstrated high classification accuracy in both online and offline experiments.
• The developed online BCI system proved viable and effective for controlling LLEs based on motor imagery.
• Eight subjects participated, validating the system's performance and potential.

Conclusions:

• The online BCI system significantly enhances patient involvement in LLE-based rehabilitation.
• The WRF-SVM algorithm offers a promising approach for precise decoding of brain signals in BCI applications.
• This technology holds substantial potential for practical applications in lower limb exoskeleton rehabilitation.