Abstract: Low precision micro-inertial gyros and accelerometers were used as elements of an inertial strapdown attitude system. Firstly, the nonlinear problem and the cause of accumulated error for inertial strapdown attitude solution were analyzed. Then according to the trigonometric function relationship between attitude angles, three variants replaced the roll and the pitch as state variants. Furthermore, with the restrained equation among state variants and the forward motion speed, one extended Kalman filter(EKF) model was built to solve the attitude angles. Using micro-inertial strapdown measurement unit MTi in the experiment to verify whether the EKF can restrain the accumulated error of the inertial attitude solution. The two results which come from EKF and from inertial attitude solution without compensation were compared, and show that the EKF algorithm can restrain the accumulated error.