Autocorrelation analysis of accelerometer signal to detect and count steps of smartphone users

Abstract

This paper proposes a reliable step counting system for pedestrians based on pattern recognition, more specifically, on autocorrelation. It uses the accelerometer signal obtained through the accelerometer sensor that is usually present in almost all smartphones and was designed to be easily integrated as a system component in an Indoor Positioning System. The system has been thoroughly tested using specific planned scenarios always with more than one hundred steps. The ground truth was carefully obtained by counting the real number of steps. The same path was travelled many times, by different users, carrying the smartphone in 5 distinct positions: in hand, in front pocket, in back pocket, all-mixed and running. Experimental results show that on average only one step is miscounted using the proposed method, regardless of the position of the smartphone.