Physiological modules for generating discrete and rhythmic movements: action identification by a dynamic recurrent...

Area of Science:

• Neuroscience
• Biomechanics
• Robotics

Background:

• Understanding the neural control of complex human movements is crucial for rehabilitation and prosthetics.
• Muscle synergies, or coordinated activation patterns, are thought to simplify motor control.

Purpose of the Study:

• To investigate the underlying control modules of muscle activations during figure-eight arm movements.
• To explore the generalization capabilities of a dynamic recurrent neural network (DRNN) in predicting hand movements from EMG signals.

Main Methods:

• Healthy subjects performed figure-eight movements in frontal and sagittal planes.
• A DRNN was trained to predict wrist movement from EMG data of seven muscles.
• The DRNN's generalization ability was tested across different movement directions and planes.

Main Results:

• A DRNN trained on all movement directions within a plane generalized better than one trained on a single direction.
• The DRNN could reproduce kinematics across planes only if trained on data from both.
• Analysis suggested three fundamental control modules for discrete-rhythmic movements.

Conclusions:

• EMG to hand movement mapping is constrained by movement dynamics and musculoskeletal anatomy.
• Discrete-rhythmic movements may emerge from a combination of co-activation and reciprocal activation modules.
• This approach offers insights into the neural basis of motor control.