Evolving Gaussian Process Autoregression Based Learning of Human Motion Intent Using Improved Energy Kernel Method of...

Area of Science:

• Robotics and Human-Computer Interaction
• Biomedical Engineering and Signal Processing

Background:

• Continuous human motion intent learning is crucial for advanced human-robot interaction and prosthetics.
• Existing methods struggle with unspecified and time-varying motion patterns, necessitating more flexible modeling approaches.

Purpose of the Study:

• To develop a robust and computationally efficient system for continuous human motion intent learning.
• To leverage Gaussian Processes (GP) and electromyography (EMG) signals for accurate prediction of human kinematics and intent.

Main Methods:

• Utilized Gaussian Process (GP) autoregression within a nonlinear autoregressive with exogenous inputs (NARX) framework.
• Employed an evolving system to capture irregular and unspecified dynamic motion features.
• Processed electromyography (EMG) signals using a novel energy kernel method to extract muscle activation and force information.

Main Results:

• Demonstrated superior flexibility in learning human kinematics through the statistical nature of GP.
• Achieved significant improvements in computational efficiency for real-time applications without compromising robustness.
• Successfully extracted muscle activation levels, muscular force, and motion intent from EMG signals.

Conclusions:

• The proposed GP-based evolving system effectively models continuous human motion intent from EMG signals.
• The energy kernel method offers an efficient and robust way to process EMG for motion intent recognition.
• This approach enables credible motion intent prediction and facilitates risk-based control systems.