Fuzzy inference system (FIS) - long short-term memory (LSTM) network for electromyography (EMG) signal analysis.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Accurate decoding of neuromuscular signals like Electromyography (EMG) is crucial for applications such as remote surgery and rehabilitation.
• Existing methods often fail to capture temporal correlations in EMG signals, limiting classification accuracy.
• Motor states (e.g., hand movements) are encoded in EMG patterns, requiring sophisticated analysis for reliable interpretation.

Purpose of the Study:

• To develop and evaluate a novel classification technique for motor states using Electromyography (EMG) signals.
• To improve the accuracy of classifying hand movements by considering temporal signal correlations.
• To provide human-interpretable results for potential use in medical devices and rehabilitation.

Main Methods:

• A hybrid approach combining a Fuzzy Inference System (FIS) with a Long Short-Term Memory (LSTM) neural network was proposed.
• Features were extracted from EMG signal patterns within a defined time window, capturing temporal dynamics.
• The system was trained and tested for classifying four distinct hand motor states: Forward, Reverse, GripUp, and RelDown.

Main Results:

• The proposed FIS-LSTM model achieved high classification accuracies: 91.3% for four-way action classification.
• Specific two-way action classifications demonstrated even higher accuracy: 95.1% for Forward/Reverse and 96.7% for GripUp/RelDown.
• The method proved effective in extracting and utilizing temporal features from EMG signals for robust motor state decoding.

Conclusions:

• The novel FIS-LSTM technique offers a significant advancement in EMG signal classification for motor state decoding.
• The approach provides accurate and human-interpretable results, suitable for real-world applications in rehabilitation and medical technology.
• This method addresses limitations of traditional techniques by incorporating temporal signal dynamics for enhanced reliability.