The human body has more muscles than Degrees of Freedom (DoF), and that leads to indeterminacy in the muscle force calculation. This study proposes the formulation of an optimization problem to estimate the lower-limb muscle forces during a gait cycle of a patient wearing an instrumented knee prosthesis. The originality of that formulation consists of simulating muscle excitations in a physiological way while muscle parameters are calibrated. Two approaches have been considered. In Approach A, measured contact forces are applied to the model and all inverse dynamics loads are matched in order to get a physiological calibration of muscle parameters. In Approach B, only the inverse dynamics loads not affected by the knee contact loads are matched. With that approach, contact forces can be predicted and validated by comparison with the experimental ones. Approach B is a test of the optimization method and it can be used for cases where no knee contact forces are available.