Spatio-spectral filters for low-density surface electromyographic signal classification

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Engineering

Background:

• Surface electromyography (EMG) signals are crucial for recognizing intended motions but are often contaminated by noise.
• Conventional time-domain feature extraction methods have limitations in accurately classifying motions from noisy EMG data.
• Developing advanced filtering techniques is essential for improving the reliability of EMG-based motion identification.

Purpose of the Study:

• To introduce and evaluate a novel spatio-spectral filter, the Common Spatio-Spectral Pattern (CSSP), for enhancing EMG signal classification.
• To assess the effectiveness of the CSSP method in improving the accuracy of wrist and hand motion recognition.
• To explore the potential of CSSP for functional prosthetic control in amputees.

Main Methods:

• Utilized a novel Common Spatio-Spectral Pattern (CSSP) filter, a classification-oriented optimal spatio-spectral filter.
• Collected low-density (six channels) surface EMG signals from five able-bodied subjects and one transradial amputee during an eight-task wrist and hand motion recognition experiment.
• Compared the classification accuracy of the CSSP method against conventional time-domain feature extraction methods.

Main Results:

• The CSSP method demonstrated substantially improved classification accuracy compared to time-domain and other methods across all able-bodied subjects.
• Cross-validation confirmed the superior performance of the CSSP method in EMG-based motion recognition.
• The CSSP method achieved better classification accuracy in the transradial amputee, indicating its effectiveness in diverse populations.

Conclusions:

• The Common Spatio-Spectral Pattern (CSSP) filter significantly enhances classification accuracy for identifying intended motions from low-density surface EMG signals.
• CSSP effectively separates discriminative information from noise, outperforming conventional methods.
• The CSSP method shows significant potential for improving functional prosthetic control by enabling more accurate motion recognition in amputees.