Abstract: Objective:To establish the models of the waist acceleration and its derivative variables to estimate the net energy expenditure as well as the speed of human movement,this study is to investigate whether antero-posterior or vertical axis is more sensitive to estimate energy expenditure,whether different acceleration data processing methods impacted accuracy of estimating the energy expenditure.Methods:Eight male college student volunteers were studied.Energy expenditure and accelerometer parameters were collected during treadmill walking.Results:The sum of two axes detrended acceleration integral value in the walking and net energy expenditure had a high linear relationship(R2=0.92,F=731.14),there was a high linear relationship between the RMS(Root-Mean-Square) of the detrended vertical acceleration and net energy expenditure too(R2=0.92,F=700.71),but the former was higher.Net energy expenditure of the integral value had a high linear correlation with the detrended vertical and antero-posterior of the acceleration integral value respectively(R2=0.91,F=632.88;R2=0.81,F=261.08);The multiple correlation coefficient between the RMS of the acceleration as well as integral values and walking net energy expenditure was low without detrended data processing.Conclusions:The value of detrended acceleration integral as well as RMS of detrended acceleration and walking net energy expenditure as well walking speed has a high linear relationship;Vertical sensitive axis is better than antero-posterior sensitive axis when estimating energy expenditure and the walking speed.