The Effectiveness of Narrowing the Window size for LD & HD EMG Channels based on Novel Deep Learning Wavelet...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Technology

Background:

• Electromyogram (EMG) signal complexity challenges prosthetic control.
• Feature extraction and window size critically impact classification accuracy.
• Deep learning models require extensive data and high computational resources.

Purpose of the Study:

• To evaluate the deep wavelet scattering transform (DWST) for low-cost feature extraction from EMG signals.
• To investigate DWST's performance on high-density (HD) and low-density (LD) EMG datasets.
• To reduce analysis window size without compromising classification performance for real-time applications.

Main Methods:

• Utilized DWST for feature extraction from EMG signals.
• Examined feature extraction from both HD and LD EMG datasets.
• Reduced analysis window size to 32ms, assessing impact on classification accuracy.
• Compared DWST performance against established feature extraction algorithms.

Main Results:

• DWST demonstrated low computational cost for EMG feature extraction.
• Feature extraction using DWST on HD and LD EMG datasets was effective.
• Minimal impact on classification performance was observed with reduced window sizes.
• The proposed DWST strategy achieved over 25% higher accuracy compared to other methods.

Conclusions:

• DWST is a computationally efficient and effective feature extraction technique for EMG signals.
• DWST offers a viable alternative to deep learning for prosthetic control applications.
• The method enables reduced window sizes, supporting real-time prosthetic system operation.