Cycle characteristics in roller skiing measured by IMU sensors =

In cross-country (XC) skiing, double poling (DP) is a widely used skiing technique in which skiers push synchronously two poles to generate propulsion. Cycle parameters as cycle time, push and recovery time, cycle speed, and cycle length play an important role in performance assessment. The use of inertial measurement units (IMUs) to assess cycle parameters have the advantages to require a simple setup and to allow measuring skiing kinematics continuously outside. The aim of the thesis was to identify cycle events (pole hits and lifts) and to calculate cycle parameters by linear acceleration and angular velocity signals collected using IMUs fixed on the skiers, and to assess errors (accuracy and precision) of the results in comparison with a gold standard system. Two expert skiers performed DP on a treadmill at three speeds and two inclinations. The participants skied using poles with force sensors embedded, which was the gold standard to detect pole cycle events, and they wore 6 IMUs placed on the right arm and forearm (one proximally and one distally), upper back (C7) and right ski. The protocol was repeated with IMUs aligned and rotated with respect body segments. Three different IMUs signals-based cycle events identification methods were developed: (1) acceleration norm method, (2) angular velocity norm method, and (3) fusion method, based on the fusion of method (1) and (2). In addition, the relationship between the linear acceleration and angular velocity mean, integral, peak and the skiing speed and terrain inclination was assessed. Finally, ski inclination was calculated with the IMU placed on the ski using two methods: angular velocity integration and accelerometer as inclinometer. The ski velocity and displacement were calculated by integrating linear acceleration along the ski longitudinal axis. The 4 sensors on the upper limb showed small median error (MD) and interquartile range errors (IQR) in identifying pole hits (MD<44 ms and IQR<30 ms) and in calculating cycle time (MD<2 ms and IQR<14 ms) with the all the three cycle events identification methods. The wrist sensor provided the best performance in calculating cycle time (MD<2 ms and IQR<4 ms) and it was the one to be suitable also to calculate push time and rest time (MD<38 ms and IQR<16 ms). The use of the signals norm in the pole events identification ensured independence from the sensors orientation. The linear acceleration mean appeared to be particularly sensitive to skiing speed variations, whereas angular velocity integral seemed more influenced by treadmill inclination. Regarding the ski inclination calculation, inclinometer-based method showed greater agreement with treadmill inclination compared to angular velocity method, which is subjected to drift. The ski velocity and displacement calculated with the ski sensor agree with the reference depending on the trial; therefore, this methodology must be improved. In conclusion, this thesis achieved the proposed goal identifying three methods to assess accurately and precisely the DP cycle temporal parameters with one IMU on the wrist and proposed a simple method to investigate the ski inclination in XC skiing DP.