Selection of sampling rate for EMG pattern recognition based prosthesis control.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Engineering

Background:

• Conventional electromyography (EMG) sampling rates are standard in myoelectric prosthesis research.
• The impact of reduced sampling rates on EMG pattern recognition accuracy remains unclear.

Purpose of the Study:

• To investigate the effect of varying EMG sampling rates on the performance of pattern recognition for multifunctional myoelectric prostheses.
• To determine an optimal sampling rate for accurate classification of upper limb movements.

Main Methods:

• EMG data were collected from able-bodied subjects and transradial amputees performing 11 classes of arm and hand movements.
• Classification accuracy was assessed at different sampling rates (1 kHz, 500 Hz, 400 Hz).
• Pattern recognition algorithms were used to identify user movement intents.

Main Results:

• A sampling rate of 500 Hz resulted in a minimal average accuracy decrease of 0.8% (able-bodied) and 2.2% (amputees).
• A 400 Hz sampling rate led to a 1.3% (able-bodied) and 2.8% (amputees) decrease in average classification accuracy.
• Accuracy was largely maintained even with reduced sampling rates.

Conclusions:

• A sampling rate between 400-500 Hz is suggested as optimal for EMG acquisition in pattern recognition-based myoelectric prosthesis control.
• Lower sampling rates can effectively capture neural control information for movement intent classification.
• This finding could lead to more efficient and potentially lower-cost myoelectric prosthesis systems.