Muscle forces can be predicted by an optimization method. Researchers use various parameters for this method. This means that the optimization method has not been established yet. The present paper examines how these parameters affect correlation coefficients between predicted muscle forces and electromyogram (EMG) to search a better combination of these parameters. The parameters investigated were (1) two moment-arms, (2) three kinds of denominators of objective function (methods to normalize predicted muscle forces), (3) five kinds of physical cross-sectional areas (PCSA) and (4) three kinds of muscle length and muscle fiber length. Three male subjects walked three times. The model we used had three joints and nine muscles of the lower leg in sagittal plane. Electrodes were placed on gluteus maximus, semitendinosus, rectus femoris, vastus medialis, gastrocnemius, soleus and tibial anterior muscle. Results showed that, in the best combination of the investigated parameters, the correlation coefficients between predicted muscle forces and EMG activities at six muscles except rectus femoris were 0.79 or more. In most cases, the coefficients at rectus femoris were 0.2 or less. Otherwise, the coefficients at rectus femoris were high (r=0.47 or 0.55), and the coefficients at hamstring and vasti muscles were lower.
View full abstract