A SVM-based adaptive stance detection method for pedestrian inertial navigation

This paper presents a foot-mounted inertial sensor system for pedestrian localization, and aims to find an adaptive stance detection method for pedestrian gait analysis. The approach is based on a Support Vector Machine (SVM) classifier, which divides the gaits into two types: walking and running. For walking, the algorithm uses two threshold conditions and a median filter to detect stance and still phases. For running, a new step detection method based on Extended Kalman Filter (EKF) is used to roughly identify every step of running at first, and then empirical formulas are summarized between the average velocity of each step and thresholds. The corrected thresholds based on empirical formulas are used in the second-round accurate stance detection. The localization accuracy for running is largely improved in this algorithm.