Position estimating fusion sensor system is proposed to control attitude of the humanoid robot. To apply vision sensor to control the attitude of the humanoid robot we should solve limitations of vision sensor which are low sampling frequency, estimation time delay and limited estimation range. Sensor fusion with MEMS gyro sensor can remove estimation time delay and increase sampling frequency. Stabilization control can remove limited estimation range by regulating of the vision sensor to initial position all the time. In this research, humanoid robot is modeled as one axis inverted pendulum which represents a leg of the humanoid robot.