An Energy Expenditure Estimation Algorithm for a Wearable System

This paper presents of an algorithm for physical activity classification and MET mapping regression model construction for a wide range of daily activities using a wearable system. The sensor system consists of several sensor modules that can be synchronized to record the accelerations of diverse motions/activities. During the measurement the accelerations of daily activities, three sensor modules are worn at the participant’s hand wrist, waist, and ankle, respectively. In addition, the participant’s chest is attached to an indirect calorimeter (Cosmed K4b2) to measure oxygen uptake to calculate actual metabolic equivalent (MET) during the experiments. The oxygen uptake for different activities is used to construct MET mapping regression models. Our experimental results shows that the average classification accuracy of five categories of physical activities is 95.33%. The average error of MET estimation without and with activity classification is -7.25×10-15±1.16 METs and 2.31×10-4±0.71 METs, respectively.