Magnetic and inertial sensors are becoming increasingly popular to measure three-dimensional (3D) orientation, because they are well suited to the ambulatory monitoring of posture and movements of subjects. This paper presents a complete implementation of the measurement and estimation of 3D orientation based on a magnetic and inertial measurement unit (MIMU) that we developed. The measurement unit was a combination of a 3D accelerometer, a 3D gyroscope, and a 3D magnetometer. A Kalman filter-based sensor fusion algorithm was proposed to implement the measurements and 3D orientation estimates. The accuracy of the orientation estimation, calculated by a sensor fusion algorithm, was assessed by comparison with a laboratory-bound optical measurement system. Several simulation experiments were executed to evaluate the performance of the measurement unit under various states, including static, periodically rotational, arbitrarily dynamic, and vibration states. Experimental results showed accurate and drift-free orientation estimates. The averaged root-mean-square errors (RMSE) of the roll, pitch, and yaw Euler angles in static state were ≤0.6°. The averaged RMSE of the three angles in dynamic state or in dynamic tests at different angular velocities were ≤2.1°, regardless of periodic rotations and arbitrary motions.