The anatomical redundancy of the human musculoskeletal system makes a unique solution for muscle forces difficult to ascertain. In this thesis, a neuromusculoskeletal model was presented as a method for resolving individual muscle forces during complex multisegment movements.

A forward solution using an EMG driven muscle model instituted at the level of the individual muscle was proposed. The equations describing the mechanical response of the muscle model were based on Hill's (1938) original work, but incorporated muscle length, velocity and excitation considerations for eccentric and concentric muscle contractions. Processed EMG represented the neural input to the muscle. A musculoskeletal model defining the skeleton, and, line of action and architecture of each of the lower limh, muscles was developed. Muscle kinematics were then calculated using the musculoskeletal model in conjunction with three dimensional cinematography. Muscle force as a function of length and level of excitation was also required as input to the model, and was established from a series of slow isokinetic calibration Contractions.

Individual muscle force profiles were predicted for selected muscles of the lower limb using two subjects and two movement conditions;​ a normal walk, and a single leg squat motion. Results using the model were validated by summing moments calculated from the predicted, muscle forces and comparing them to net joint moments calculated from limb kinematics and ground reaction forces using link segment mechanics.

The moment curves matched closely in shape. The correlation between moments derived from the two approaches ranged from r=.72 to r=.97 for the gait trials. The correlations were also high for the squat movement. The R.M.S. difference between moment curves over one walking stride was about 12 N.m at the ankle, 25 N.m at the knee and 18 N.m at the hip. R.M.S. differences were slightly higher for the squat trials, ranging from 13 to 27 N.m. Expressed as a percentage of the R.M.S. of the moments calculated using the inverse dynamics solution, the differences ranged from 23 to 34 percent at the ankle, 29 to 79 percent at the knee and 59 to 103 percent at the hip. Thee results were greater than R.M.S. differences of:​ 7 N.m at the ankle and 11 N.m at the knee, calculated from the model predictions of Olney and Winter (1985), but less than a R.M.S. difference of 23 N.m reported at the ankle (Hof et al., 1983) using the muscle model of Hof, and Van Den Berg (1981). No studies could be found for comparing the results at the hip. It should be noted that the results of Olney and Winter (1985) were achieved by optimizing model parameters until a best fit to the moment-curve calculated for the walking trial was realized. The approach of this thesis differs in that the model parameters were not optimized to fit the moment curves.

The nature of the moment curve differences suggested that the equation formalism of the muscle model was essentially correct, but that the constants in the model may not have been optimal. The results of the model prediction could be improved, with a more accurate EMG calibration and a more accurate placement of surface markers identifying the bony landmarks used to determine the spatial location of segments and joint centres.

The results from this research support the feasibility of using the neuromusculoskeleftal modelling approach proposed, as a potential solution to the indeterminancy problem, thereby giving a unique solution to muscle forces involved in normal human movements.