Signal Processing for Estimating Energy Expenditure using Triaxial Accelerometers

Abstract

PURPOSE: The purpose of this study was to examine possibility of integral method as signal processing for estimating energy expenditure (EE) using triaxial accelerometers. METHODS: 17 healthy, nontrained males (age: 27.23±2.18, range 24-31yr, VO2max: 45.96±5.93 ml.kg/min) participated in this study. All subjects completed a submaximal treadmill GXT according to modified bruce protocol. A device that is able to measure the variation in values from the triaxial acceleration sensors during exercise and to save those values was designed and fabricated in this study. Accelerometers were worn on the right ankle. The data fluctuation on the three axes during exercise was sampled at 100Hz and the fluctuation on each axis (x, y, z) was saved separately in the micro SD memory. Accelerometer signal and EE using respiratory gas analysis were collected during an incremental exercise test on treadmill at the same time. Integral and count method were evaluated as signal processing for estimating EE using triaxial accelerometer. Pearson correlation analysis and multiple regression (SPSS 15.0 for windows) were used to identify an appropriate EE predictor. The significance level was p<0.05. RESULTS: Correlation coefficient between accelerometer integral value and EE (r=0.918, p<0.001) was larger than one between accelerometer count value and EE (r=0.713, p<0.001). AS the predictor of multiple regression equation, the acceleration signals, hours of exercise(r=0.883, p<0.01), and weight (r=0.154, p<0.01) variables were selected. Multiple linear regression of EE estimation using integral and count method generated the following result (R2=0.935, SEE=8.39 vs R2=0.891, SEE=10.27, respectively). CONCLUSIONS: Findings from this study indicate that a proposed approach using accelerometer integral method can provide acceptable estimation of EE during incremental treadmill exercise. And accelerometer integral method is more effective than accelerometer count method in estimation of EE. An integral method will permit more accurate prediction of EE under free-living conditions as well as during exercise.