Running patterns analysis using inertial measurement units

Abstract

Since more than ten years, running is experiencing a boom accompanied by a proportional explosion of research in that field. Quantitative analysis is of critical interest for both clinical and sport science fields. Inertial measurement units (IMU) are now widely considered as an economical alternative for long assessment in real conditions. Detecting the instant at which the foot strikes the ground (initial contact) and the instant at which it leaves the ground (terminal contact) gives an access to many useful temporal parameters like the stance, the swing or the stride duration. In this thesis, an original algorithm to detect the main running temporal events is presented. This algorithm allows to detect initial contacts with a precision (median $\pm$ interquartile range IQR) of 5 $\pm$ 10 and $0 \pm 5$ for terminal contacts. The absolute errors in the stride and stance durations are of -1.56 $\pm$ 3\% and 0 $\pm$ 1.32\%.