Description

The standard approach to experimentally measure human motion, for instance, gait, is to attach reflective markers to the trunk, legs and feet of the test subject. The 3D coordinates of these markers are subsequently measured by a camera system. An important task in the processing of this marker data is to convert the marker coordinates into joint angles, velocities and accelerations of the various body segments (e.g., trunk, upper leg, lower leg and foot), based on a model of the test subject's musculoskeletal system.

The main drawback of current state-of-the-art nonlinear least-squares methods is that they result in very noisy joint acceleration estimates. Recently, however, a new nonlinear Kalman-smoother based method has been developed at the K.U.Leuven (Belgium) which substantially improves upon these methods. This algorithm is currently implemented in Matlab and rather slow.

The goal of the practical is to implement the Kalman smoother in C++, so as to substantially decrease the required computational time. If time permits, also a simple GUI can be implemented. Experimental data is provided by K.U.Leuven.

Requirements

• A good knowledge of C++ is mandatory.
• Knowledge of Kalman filtering is a plus but not necessary

Size of the project

This project is suited for one or two students, as an advanced practical.

Contact