Synergistic control of forearm based on accelerometer data and artificial neural networks

Area of Science:

• Biomechanics
• Robotics
• Rehabilitation Engineering

Background:

• Upper limb movement relies on complex coordination between the upper arm and forearm.
• Restoring motor function after injury, such as spinal cord injury or stroke, is a significant challenge.
• Current rehabilitation methods may benefit from advanced predictive control systems.

Purpose of the Study:

• To model upper limb reaching movements using a two-link mechanism.
• To develop and train an artificial neural network (ANN) to predict forearm motion from upper arm trajectories.
• To evaluate the ANN's generalization capabilities for various movements and subjects.

Main Methods:

• Modeled reaching tasks as a two-link mechanism.
• Estimated upper arm and forearm motion from angular accelerations using dual-axis accelerometers.
• Trained a radial basis function artificial neural network (RBF-ANN) on reaching synergies from able-bodied individuals.
• Tested the RBF-ANN's predictive accuracy for forearm motion based on new upper arm trajectories.

Main Results:

• The trained RBF-ANN predicted forearm motion from similar upper arm trajectories with high correlation (0.9149-0.941).
• Prediction accuracy was significantly lower for dissimilar movements (0.0316-0.8302).
• The algorithm demonstrated successful generalization over similar motions and subjects.

Conclusions:

• The proposed ANN algorithm can predict forearm kinematics from voluntary upper arm movements.
• This methodology is suitable for restoring upper limb function in individuals with forearm motor disabilities.
• The approach is applicable to upper-limb orthotics, functional electrical stimulation, and rehabilitation monitoring, particularly for spinal cord injury and stroke patients.