Hand Movement Classification Using Burg Reflection Coefficients.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Electromyography (EMG) signal classification is crucial for clinical diagnosis and prosthetic device control.
• Accurate and hardware-implementable algorithms are essential challenges in EMG signal classification.
• Existing methods may omit useful features, impacting machine learning model performance.

Purpose of the Study:

• To propose an effective electromyography signal classification method.
• To improve classification accuracy using signal modeling and time-domain characteristics.
• To reduce feature dimensionality while maintaining high classification performance.

Main Methods:

• Signal processing and feature extraction techniques were applied to EMG signals.
• Burg reflection coefficients were utilized for creating learning and classification patterns.
• Feature selection algorithms were employed to identify the most relevant attributes.
• Hand movement identification was used as a benchmark for classification.

Main Results:

• The proposed method using Burg reflection coefficients achieved competitive classification rates compared to traditional time-domain features.
• Feature selection algorithms significantly improved classification performance.
• A high classification rate of up to 100% was attained with low pattern dimensionality.
• The approach demonstrated effectiveness in hand movement identification.

Conclusions:

• The proposed signal processing and feature extraction method, incorporating Burg reflection coefficients and feature selection, enhances EMG signal classification accuracy.
• This approach offers a promising solution for developing efficient and accurate control systems for prosthetic devices.
• The method achieves high performance with reduced feature sets, suitable for hardware implementation.