Surface EMG Pattern Recognition Using Long Short-Term Memory Combined with Multilayer Perceptron.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) signals are crucial for prosthetic limb control.
• Accurate motion classification from sEMG is essential for intuitive and effective prosthetic function.
• Existing pattern recognition methods face challenges in capturing complex sEMG signal dynamics.

Purpose of the Study:

• To develop and optimize a hybrid neural network model for enhanced sEMG-based motion classification.
• To leverage the strengths of Long Short-Term Memory (LSTM) and Multiplayer Perceptron (MLP) networks for feature learning.
• To improve the accuracy and robustness of prosthetic control systems.

Main Methods:

• A hybrid pattern recognition model combining LSTM and MLP networks was proposed.
• The LSTM network was utilized to capture temporal dependencies in sEMG signals.
• The MLP network focused on static characteristics, and network architecture (width, depth, dropout) was optimized.
• Performance was evaluated using the Ninapro database.

Main Results:

• The proposed hybrid model demonstrated superior motion classification accuracy compared to existing techniques.
• Integration of LSTM and MLP effectively captured both dynamic and static features of sEMG signals.
• Optimized network architecture contributed to achieving the best classification outcomes.

Conclusions:

• The combined LSTM-MLP model offers a significant advancement in sEMG-based motion classification for prosthetic control.
• This approach provides a more comprehensive feature representation, leading to improved accuracy.
• The findings suggest a promising direction for developing next-generation intelligent prosthetic systems.