Experiment Design Considerations for Estimating Energy Expenditure during Wheelchair Propulsion

Abstract

We propose a methodology for estimating energy expenditure (EE) during wheelchair propulsion. The method is based on measured physiological and kinematic signals from wearable sensor devices in an experimental setup design. More specifically, we have developed regression models based on features extracted from heart rate, acceleration and gyroscope data collected during nine experiment stages with twenty participants. Support Vector regression and Gaussian process regression methods were implemented to provide an estimate of EE for each participant during the experiment. Extensive cross validation techniques were applied to evaluate the performance of the proposed models and investigate the necessity of personalizing the algorithms based on personal characteristics.