Can MIMUs positioned on the ankles provide a reliable detection and characterization of U-turns in gait?

The detection and characterization of U-turns during locomotion using Magneto-Inertial Measurement Units (MIMU) with the goal of segmenting the walking trial into straight walking bouts and turns is an open issue currently under investigation. Typically, a MIMU located on the lower back or trunk is used to this purpose and turns onset timing and duration are determined. The aim of this study was to assess if an existing method could be satisfactorily applied to signals recorded from MIMUs near the ankles. Additionally, a method is proposed with the aim of limiting the differences of its output from that of the existing method guaranteeing high robustness with respect to the MIMU location. The analysis was conducted on data recorded from healthy elderly subjects and patients with Parkinson's disease walking at two different speeds. The existing method applied to signals from the MIMU near the ankles could detect the same number of U-turns as the original. However, their onset and duration were often more than 200ms away from those obtained with the original method. Similar results were obtained with the proposed method, showing some limitations in part related to the heuristic threshold employed. However, the proposed method demonstrated a superior robustness with respect to the MIMU location. Overall, the proposed method appears to be a good starting point for the definition of a more stable and robust method for U-turn detection and characterization from signals recorded from MIMUs near the ankles