Position-independent gesture recognition using sEMG signals via canonical correlation analysis

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Signal Processing

Background:

• Surface electromyogram (sEMG) signals are crucial for gesture recognition in human-computer interaction.
• Variations in arm position during gesture execution lead to inconsistent sEMG signals, challenging traditional recognition methods.
• Existing solutions either reduce accuracy by ignoring position differences or increase training complexity by classifying gestures and positions separately.

Purpose of the Study:

• To develop a novel framework, Position Independent Canonical Correlation Analysis (PICCA), to address sEMG signal variability caused by different arm positions.
• To explore intrinsic position-independent (PI) characteristics of sEMG signals for improved gesture recognition.
• To enhance classification accuracy in both user-dependent and user-independent scenarios without significant increases in training burden.

Main Methods:

• Proposed a novel framework, PICCA, leveraging Canonical Correlation Analysis (CCA) to extract position-independent features from sEMG signals.
• Utilized a predefined expert set to map both training and testing datasets into a unified style via CCA.
• Evaluated the framework on 13 gestures across 3 different arm positions.

Main Results:

• PICCA demonstrated significant improvements in classification rates compared to methods without CCA.
• Achieved 28.52% and 44.19% promotion in classification rates for user-dependent and user-independent manners, respectively.
• Maintained acceptable training burdens while enhancing accuracy.

Conclusions:

• The PICCA framework effectively extracts position-independent characteristics from sEMG signals, improving gesture recognition accuracy.
• This approach offers a viable solution for robust myoelectric interfaces, overcoming the challenge of arm position variability.
• The findings facilitate the practical implementation of more reliable and accurate myoelectric control systems.