The estimation of the energy expenditure during physical activities has been considered as an important subject and studied consistently for a long time. Nineteen male healthy subjects participated and were required to walk and run on the treadmill with the gas analyzer and a tri-axial accelerometer. The tri-axial accelerometer was attached on the point of middle of left and right posterior superior iliac spine (PSIS). Walking speed were 3.0, 4.0, 5.0 and 6.0 km/h and running speed were 7.0, 8.0 and 9.0 km/h respectively. One trial consists of three periods; ramp-up (1 minute), practice for walking or running (5 minutes), and cool-down (1 minute). After being removed first two minutes during practice, which was expected 'steady-state' for minimizing the errors of unstable energy expenditure, the remaining data was then used for analysis. To find a linear relationship between energy expenditure from the gas analyzer and that of accelerometer data, physical activity was calculated by the integration of the accelerometer. The linear relationship between the measured and estimated values were excellent in both walking (r = 0.988) and running (r = 0.895). These results would be expected to apply to health management product such as preventing fatness system.