Contraction-based variations in upper limb EMG-force models under isometric conditions.

Area of Science:

• Biomechanics
• Neuromuscular modeling
• Human movement analysis

Background:

• Accurate estimation of joint torques is crucial for understanding and predicting human movement.
• Existing models often struggle to precisely capture the complex relationship between muscle activation and joint torque.
• Surface electromyography (EMG) provides valuable insights into muscle activation patterns.

Purpose of the Study:

• To enhance the accuracy of elbow joint torque estimation using a previously developed model.
• To investigate the impact of different dataset grouping strategies on model performance.
• To gain deeper insights into muscle behavior during dynamic movements.

Main Methods:

• A model mapping joint kinematics and estimated muscle activation to net elbow joint torque was trained.
• Four distinct dataset grouping methods were employed: within trials, across trials, within force levels, and across force levels.
• Model performance was evaluated based on average error metrics.

Main Results:

• Average evaluation error ranged from 5% to 15%.
• The lowest error was achieved when training datasets were grouped within specific force levels.
• Further error reduction was observed when training datasets were grouped across multiple data collection trials.

Conclusions:

• Dataset grouping significantly impacts the accuracy of elbow joint torque estimation.
• Grouping data by functional requirements, such as force levels and across trials, leads to more precise models.
• This approach allows for the development of tailored models that better reflect muscle behavior and improve biomechanical predictions.