In IMUs, three sensors, acceleration, gyroscope, and magnetic, are implemented. Data get from these sensors are used to acquire posture and position of each sensor. Theoretically, from a simple integration of an acceleration and an angular velocity, position and posture could be calculated. The IMU is applied to many fields due to its user friendly structure. Human motion detection is one of interesting applications. However, small error in the measured data are also integrated, hence position error and posture error, which is also called drift, occur. The purpose of this paper is to estimate position and posture using IMUs without human model. Two IMUs are used to estimate posture and position. For posture estimation, complementation of posture angle are done from an 'angular velocity' and a 'gravity acceleration'. For position estimation, complementation by relative relationship between 2 sensors has been proposed. This is the novel part of this paper. Two ways of position complementation algorithm are proposed. One method is complementation by Kalman Filter, and the other is complementation by error estimation. Error estimation is based on a theory of mode transfer between relative space and world space. The validation of proposed method is done by experiment.