An efficient approach for EMG controlled pattern recognition system based on MUAP identification and segregation.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Electromyography (EMG) signal processing is crucial for controlling external devices.
• Surface electrodes capture MUAPs from multiple muscles, complicating analysis.
• Efficient feature extraction and classification are key to accurate EMG-based control.

Purpose of the Study:

• To develop a novel signal processing approach for EMG pattern recognition using MUAP decomposition and segmentation.
• To improve the efficiency and accuracy of EMG-based control systems.
• To evaluate the performance of the proposed method using various machine learning classifiers.

Main Methods:

• A new algorithm for MUAP identification and segmentation based on primary MUAP waveshapes and correlation scores.
• Noise elimination and active muscle signal separation using a determined noise margin.
• Feature extraction from variable-width segments (110-200 ms) and classification using LDA, kNN, DT, and RF models.

Main Results:

• The proposed MUAP-based segmentation achieved a 20-50% reduction in segmentation width compared to conventional methods.
• kNN and DT classifiers demonstrated superior performance over LDA and RF.
• Maximum precision and recall reached 100%, with a maximum accuracy of 98.56%.

Conclusions:

• The novel MUAP decomposition and variable-width segmentation approach significantly improves EMG signal processing accuracy.
• This method offers a 5% to 15% increase in accuracy compared to constant window segmentation, even with reduced segmentation widths.
• The findings suggest a more efficient and accurate pathway for EMG-based control systems.