Prediction of energy consumption according to physical activity intensity in daily life using accelerometer

Abstract

This research aims to accurately predict energy consumption according to physical activity intensity in daily life using an accelerometer. To derive a simple but accurate equation for prediction of correlations between acceleration information and energy consumption according to physical activity intensity, the following practices have been undertaken in this research. First, an experiment has been conducted with accelerometers attached at the waist and wrist. Second, 13 motions in daily life with different physical activity intensities were performed. The experiment was conducted on 20 healthy persons using a respiratory gas analyzer for measurement of actual energy consumption. As a predictive model for energy consumption, a multiple regression equation was developed using accelerometer data and physical information about the subjects. For comparison between single accelerometer attached at the waist and two accelerometers attached at the wrist and waist, the correlation between actual energy consumption and accelerometer output for the former was 0.911 and that for the latter was 0.914, which is not significantly different. This result implied that one accelerometer attached at the waist is practically efficient in terms of prediction of energy consumption. As a result of comparison between a single regression equation without motion classification and several regression equations with inclusive motion classification, it was found that several regression equations showed smaller errors (0.64 vs 0.18). Thus for accurate and practical prediction of energy consumption, it is recommended to use several regression equations with inclusive motion classification and single accelerometer attached at the waist.