A noise-suppressing neural network approach for upper limb human-machine interactive control based on sEMG signals

Area of Science:

• Rehabilitation Medicine
• Human-Computer Interaction
• Robotics

Background:

• Upper limb rehabilitation robots are crucial for active training.
• Accurate motion intention recognition is vital for effective human-robot interaction in rehabilitation.
• External disturbances during training pose safety risks.

Purpose of the Study:

• To improve the accuracy and real-time recognition of active motion intention using surface electromyogram (sEMG) signals.
• To design robust human-machine interaction controllers that eliminate external disturbances for safe rehabilitation.
• To enhance the safety and comfort of upper limb rehabilitation robots.

Main Methods:

• Proposed a fuzzy neural network method for active motion intention recognition based on human upper limb sEMG signals.
• Developed two types of human-machine interaction controllers: a zeroing neural network controller and a noise-suppressing zeroing neural network controller.
• Utilized numerical experiments to validate the proposed methods.

Main Results:

• The fuzzy neural network demonstrated improved real-time recognition accuracy of active motion intention.
• The developed controllers effectively eliminated external disturbances, ensuring a safe training environment.
• The proposed approaches proved feasible and effective in numerical experiments.

Conclusions:

• The fuzzy neural network method enhances active motion intention recognition for upper limb rehabilitation robots.
• The novel controllers contribute to a safer and more comfortable rehabilitation training experience by mitigating disturbances.
• This research advances the field of human-machine interaction control in robotic rehabilitation.