Accurate estimation of the gait parameters is one of the key technologies in both pedestrian navigation system and pedestrian health monitoring. A calculation algorithm of pedestrian dynamic step length is proposed in this paper, which can solve the current problems of low accuracy and weak adaptability in pedestrian step length estimation algorithm. First, the pedestrians gait feature is decomposed, the improved zero velocity detection is used to determine the state of pedestrian movement. The Kalman filtering technology is utilized to reduce the influence of the cumulative error of the inertial sensors. Then the filtered and coordinate transformed acceleration is doubly integrated so as to get the trajectories of pedestrians. The algorithm is validated by experiments using a MTI700 inertial sensor module. The traversed distance of the subjects were calculated with less than 3.0% error with respect to actual walking distance. The algorithm is of high precision and strong adaptation compared with the existing algorithms of pedestrian dynamic step length, it is the first time that the trajectory of pedestrian been calculated and it has great prospect in the application of pedestrian navigation field.