Actor-Critic Reinforcement Learning Based Algorithm for Contaminant Type Identification in Surface Electromyography...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Surface electromyography (sEMG) signals are crucial for neuromuscular studies but are susceptible to various interferences.
• Accurate detection and mitigation of contaminants like electrocardiography (ECG) interference, movement artifacts, power line interference, and noise are essential for reliable sEMG analysis.

Purpose of the Study:

• To develop and evaluate an innovative Reinforcement Learning (RL) based approach for detecting contaminant types in sEMG signals.
• To minimize the impact of detected contaminants on sEMG signal quality.

Main Methods:

• An agent-environment model was designed, with the environment representing muscle electrical activity and the agent acting as a controller.
• The state was defined by six signal features, and actions involved applying filters to reduce interference.
• The Actor-Critic method was employed for the RL agent's learning process.

Main Results:

• The RL agent achieved an average accuracy of 92.96% in detecting four distinct contaminant types in an off-line experiment.
• The approach demonstrated effectiveness in identifying electrocardiography (ECG) interference, movement artifact, power line interference, and additive white Gaussian noise.

Conclusions:

• The proposed RL-based method offers a promising solution for automated contaminant detection and mitigation in sEMG signals.
• This approach can significantly enhance the reliability and accuracy of sEMG data analysis in various biomedical applications.