EPOC aware Energy Expenditure Estimation with Machine Learning

In 2014, 39 % of adults were overweight, and 13 % were obese. Clearly, knowing exact energy expenditure (EE) is important for sports training and weight control. Furthermore, excess post-exercise oxygen consumption (EPOC) must be included in the total EE. This paper presents a machine learning-based EE estimation approach with EPOC for aerobic exercise using a heart rate sensor. On a dataset acquired from 33 subjects, we apply machine learning algorithms using Weka machine learning toolkit. We could achieve 0.88 correlation and 0.23 kcal/min root mean square error (RMSE) with linear regression. The proposed model could be applied to various wearable devices such as a smartwatch.