Using the electromyogram to anticipate torques about the elbow.

Area of Science:

• Biomechanics
• Neuroscience
• Human Movement Science

Background:

• Processed electromyogram (EMG) signals precede muscle mechanical tension by 50-100 ms.
• This electromechanical delay offers potential for anticipating muscle activity and joint torque.

Purpose of the Study:

• To investigate the efficacy of surface EMG in estimating joint torque at future time points.
• To compare advanced EMG processing techniques with conventional filtering methods for torque estimation.

Main Methods:

• Recorded surface EMG from biceps and triceps muscles of 54 subjects during isometric contractions.
• Employed higher-order Finite Impulse Response (FIR) models with advanced processing (whitening, multi-channel EMG).
• Compared FIR models against conventional Butterworth filtering for estimating elbow joint torque.

Main Results:

• Advanced FIR models achieved a minimum torque estimation error of 5.48 ±2.21% MVC(F) within a 0-60 ms time advance.
• Conventional Butterworth filtering resulted in a statistically inferior minimum error of 6.90 ±2.39% MVC(F) at a 60 ms advance.
• Estimation error increased with larger time advances for all methods, but advanced models maintained lower error over a wider range.

Conclusions:

• Higher-order FIR models with advanced EMG processing significantly improve joint torque estimation accuracy.
• These advanced models eliminate the need for explicit electromechanical delay selection.
• The findings support the use of advanced EMG processing for reliable prediction of muscle mechanical activity.