Input error analysis of an EMG-driven muscle model of the plantar flexors

EMG is a useful tool for quantifying muscle forces and studying motor control strategies. However, the relationship between EMG and muscle force is not trivial, and depends in part on muscle dynamics. This work has the following objectives: the first, to find muscle excitations and partial joint torque contribution patterns in isometric plantar flexions, considering low and medium/high contractions. The second, to correlate such patterns with an EMG-driven muscle model error, indirectly assessed by the associate joint torques. Individual muscle contributions were calculated using the model driven by the measured EMG and compared to the total joint torque from dynamometric measurements. Thirteen young males performed a protocol with low and medium/high intensities contractions. Input functions were the normalized EMG of each triceps surae and tibialis anterior muscles. RMS error was calculated between the measured and estimated torque curves. The trends observed were: the order of individual muscle contributions to the total torque (SOL, GM, GL) was different from the order of the contraction intensities (GM, SOL, GL); the model was more accurate for medium/high contractions; the worst estimations occurred when excitation input signals found from EMG were underestimated. Possible causes for such errors and improvement suggestions are addressed.