A two-dimensional matrix image based feature extraction method for classification of sEMG: A comparative analysis...

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Rehabilitation Technology

Background:

• Computer mice are essential for human-computer interaction but inaccessible to individuals with finger disabilities.
• Surface electromyography (sEMG) reflects neuromuscular activity and can be monitored non-invasively.
• sEMG classification offers a potential solution for controlling assistive devices for the physically disabled.

Purpose of the Study:

• To develop an innovative method for extracting finger motion-generated sEMG signals.
• To apply novel features for accurate sEMG classification.
• To enable computer mouse operation for patients with physical finger impairments.

Main Methods:

• A window-based acquisition method was used to extract sEMG signal samples.
• A novel 2D matrix image-based feature extraction technique was employed, differing from traditional time/frequency domain methods.
• Machine learning classifiers (SVM, KNN, RBF-NN) were utilized on a GPU to classify extracted sEMG feature maps.

Main Results:

• All employed machine learning classifiers demonstrated effective identification and classification of sEMG samples.
• The Support Vector Machine (SVM) classifier achieved an exceptional accuracy of 100% in classifying sEMG signals.
• The novel feature extraction method proved effective in enhancing signal energy for classification.

Conclusions:

• The developed signal separation and feature extraction method is efficient and convenient for acquiring finger sEMG.
• This new approach effectively extracts features by appropriately amplifying signal energy.
• Classical machine learning classifiers performed well with the novel features, validating the method's efficacy.