A cepstrum analysis-based classification method for hand movement surface EMG signals.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Effective processing of surface electromyogram (sEMG) signals is crucial for controlling robotic and prosthetic limbs.
• Hand amputees require reliable methods for actuating exoskeleton hands based on biological signals.

Purpose of the Study:

• To propose and validate a cepstrum analysis-based method for classifying basic hand movement sEMG signals.
• To evaluate the accuracy and efficiency of this method for prosthetic hand control.

Main Methods:

• Utilized cepstrum analysis, specifically mel-frequency cepstral coefficients (MFCCs), to extract features from time-domain sEMG signals.
• Employed a generalized regression neural network (GRNN) for classifying the extracted features corresponding to basic hand movements.
• Quantified feature discrimination ability using the Kruskal-Wallis statistical test.

Main Results:

• Achieved an average accuracy of 99.34% for individual subjects and 99.23% for a collective dataset in classifying hand movements.
• Demonstrated superior classification accuracy compared to most previous studies.
• The cepstrum-based features showed significant discrimination ability.

Conclusions:

• The proposed cepstrum analysis method is effective and applicable for classifying sEMG signals of hand movements.
• The method offers high accuracy and efficient model training due to the non-iterative nature of the GRNN.
• This approach holds promise for improving the control of robotic and prosthetic exoskeleton hands for amputees.