In the inertial navigation system, the pedestrian dead reckoning (PDR) is very important in the position calculation, and the accuracy of the step count statistics directly affects the positioning accuracy of the pedestrians. Aiming at the effect of pseudopeaks in the conventional peakthreshold detection method, a peakdouble threshold step detection algorithm based on wearable MIMU is proposed. In the course of traveling, the peak detection is carried out on the combined acceleration after window function filtering and the peak is defined by the double threshold. The peak is determined as an effective peak when the peak satisfies a high threshold, and the step counting is successful when the low threshold occurs between the connected peaks, thereby reducing the influence of pseudopeak on the step counting and realizing the accurate detection of the number of steps. The experimental results show that the step counting accuracy of the pedestrian in multimotion mode is more than 98% when the MIMU is worn in various positions.