Cardinality and Short-Term Memory Concepts based Novel Feature Extraction for Myoelectric Pattern Recognition.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Traditional Electromyogram (EMG) feature extraction methods limit clinical translation due to quality issues.
• Deep learning models offer improved feature representation but face challenges with high computational costs and large data requirements.
• There is a need for efficient EMG analysis methods balancing performance and computational efficiency.

Purpose of the Study:

• To develop a novel method for extracting spatial-temporal dynamics of EMG signals by combining traditional feature extraction with Long Short-Term Memory (LSTM) concepts.
• To address the limitations of current EMG analysis methods, focusing on reduced computational cost and enhanced classification performance.
• To validate the proposed method's efficacy using diverse EMG datasets.

Main Methods:

• Integration of memory concepts from LSTM models to capture short-term temporal dependencies in EMG signals.
• Utilization of cardinality for logical combinations of spatially distinct EMG signals as a feature extraction technique.
• Development of a computationally efficient method for spatial-temporal EMG feature extraction.

Main Results:

• The proposed method demonstrates significantly enhanced myoelectric pattern recognition performance compared to existing literature methods.
• Classification accuracies reached up to 99% on validated EMG databases.
• The method achieved low computational costs while maintaining high performance.

Conclusions:

• The novel approach effectively extracts spatial-temporal dynamics from EMG signals, overcoming limitations of traditional methods.
• The combination of traditional feature extraction and LSTM memory concepts offers a computationally efficient and high-performing solution for EMG analysis.
• This method shows strong potential for clinical implementation in myoelectric pattern recognition applications.