An approach to estimating caloric expenditure during exercise activity using non-invasive Kinect camera

Estimating Caloric Expenditure is an important problem for improving exercise performance and adherence leading to improved self-management of personal wellness. Current approaches rely on marker-based systems or laboratory grade instruments that must be attached to the participants. In order to enable participants to perform exercises in their home and work environments, it is important to explore how inexpensive, non-invasive devices can be used to estimate caloric expenditure. This thesis presents an approach to estimating caloric expenditure that is based on the Microsoft Kinect camera.;In the literature, Ground Reaction Forces (GRF) are used as basis to estimate caloric expenditure; these forces were, however, measured directly using an expensive force plate system. It is shown that the 3D joint location data that are provided by the skeleton tracking algorithm for a Kinect camera can be used to estimate GRF with about 2 % accuracy for a class of exercises. These results also demonstrate that the GRF estimated from the data obtained via a Kinect camera is comparable with that estimated using more expensive marker-based systems such as the Vicon. It is shown that the data from the Kinect camera can also be used to identify different body segments of a participant. However, more sophisticated algorithms are required to accurately estimate caloric expenditure. In the future, this approach can be extended to improve the accuracy of the estimates and also consider a larger set of exercises.