MEMS (Micro-Electromechanical Systems) based IMU (Inertial Measurement Unit) including accelerometers and gyroscopes is widely used for various applications such as INS (Inertial Navigation System), pose estimation devices and others for many industries such as toy, medical, automotive and military industry. But MEMS sensor chip originally has bias and sensitivity errors from manufacturing, and there is also axis misalignment when mounting a MEMS chip on IMU PCB layer. These error factors cause inaccuracy measurement results and non-linear measurement characteristics of IMU. In this paper, accelerometers and gyroscopes are mathematically modeled based on these error factors including bias, sensitivity, coning angle and azimuth angle. Calibration procedures for accelerometers and gyroscopes are formulated using nonlinear Gauss-Newton regression logic. The effectiveness of the proposed calibration procedures are proven by simulation and experiment using high accuracy 2-axis rotational gimbal motion system.