Data-based parametric biomechanical models for cyclic motions

We present a method to convert motion capture data and anthropometric statistics into parametric biomechanical models of cyclic motions, such as walking, cycling and running. The motivation is ease of modelling and the desire to make models prospective. We have developed a data processing pipeline, which precompiles a large amount of motion capture trials into a parametric model relying on the correlations between the input variables. The compilation converts optical motion capture data into anatomical joint angle variations and anatomical body dimensions. Finally, a quadratic programming method with a closed-form solution is developed to predict motion patterns meeting subject-specific requirements. The method is demonstrated on running models, and we conclude that the method can facilitate new uses of biomechanical models.