Lower Limb Motion Recognition Based on sEMG and CNN-TL Fusion Model.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Technology

Background:

• Accurate classification of lower limb movements is crucial for developing effective rehabilitation and assistance devices.
• Existing methods using surface electromyography (sEMG) have limitations in classification accuracy.

Purpose of the Study:

• To propose and evaluate a novel fusion recognition model, CNN-Transformer-LSTM (CNN-TL), for enhanced lower limb movement classification using sEMG data.

Main Methods:

• Collected sEMG data from 20 subjects performing four distinct lower limb movements: walking upstairs, downstairs, on a level surface, and squatting.
• Preprocessed sEMG data and extracted time and frequency domain features.
• Developed and compared the CNN-TL model against CNN-LSTM, CNN, and Support Vector Machine (SVM) models.

Main Results:

• The CNN-TL model achieved higher classification accuracy compared to other models.
• Specifically, CNN-TL outperformed CNN-LSTM by 3.76%, CNN by 5.92%, and SVM by 14.92%.

Conclusions:

• The proposed CNN-TL model demonstrates superior performance in classifying lower limb movements based on sEMG signals.
• This fusion model offers an effective approach for advancing motor function in rehabilitation and assistance technologies.