Towards practical energy expenditure estimation with mobile phones

Regular physical activity plays a significant role in reducing the risk of obesity and maintaining people´s health conditions. Among all the physical activities, walking is a commonly recommended intervention for combating lifestyle diseases. The capability to accurately measure the energy expenditure of walking provides foundations to base the corresponding intervention. In this paper, we develop a set of signal processing and statistical pattern recognition techniques to estimate energy expenditure of walking in real-life settings using mobile phones. We examine the robustness of our proposed techniques to variations in location on the human body and across body types. We show that our proposed techniques can estimate step frequencies for three common locations of phone usage and achieve promising energy expenditure estimation accuracy with limited training data.