Deep Learning-Based Myoelectric Potential Estimation Method for Wheelchair Operation.

Area of Science:

• Sports Science
• Biomedical Engineering
• Rehabilitation Engineering

Background:

• Wheelchair sports competitiveness is increasing, requiring advanced performance evaluation methods.
• Electromyography (EMG) is valuable for muscle activity analysis but impractical for real-world competition due to equipment complexity and movement restrictions.
• There is a need for non-invasive, movement-unrestricted methods to measure myoelectric potential in wheelchair athletes.

Purpose of the Study:

• To develop and validate a deep learning-based method for estimating myoelectric potential in wheelchair sports.
• To enable accurate muscle activity assessment in competitive environments without hindering athlete movement.
• To provide a tool for enhanced performance evaluation in wheelchair athletics.

Main Methods:

• Developed a deep learning model integrating camera images of wheelchair movements and inertial sensor data.
• The model estimates surface myoelectric potentials based on input data.
• Validated the method by comparing estimated myoelectric potentials with EMG measurements in seven subjects during wheelchair work.

Main Results:

• The developed method demonstrated a significant correlation (correlation coefficient ≥ 0.5 at p < 0.1%) between estimated and measured myoelectric potentials.
• The in-subject model showed reliable estimation of muscle activity.
• The technique successfully estimated myoelectric potential without limiting the physical movement of wheelchair athletes.

Conclusions:

• The proposed method offers a viable solution for estimating myoelectric potential in dynamic wheelchair sports settings.
• This approach can be applied to objective performance evaluation and training optimization for wheelchair athletes.
• The non-restrictive nature of this technique opens new avenues for research and application in adaptive sports.