Rebecca Lennartz

Advisors
Arash Khassetarash PhD. (University of Calgary), Prof. Dr. Benno Nigg, Prof. Dr. Björn Eskofier

Duration
05 / 2022 – 11 / 2022

Abstract
The fit and design of sporting equipment have a large influence on athletes’ comfort, performance, and risk of injury. For example, the design of skate blades can improve the skating speed of ice hockey players by up to 1,3 % [1] or the ice hockey stick shaft stiffness can significantly influence the puck velocity in wrist shots [2]. In a fast, high-impact sport such as ice hockey, the equipment plays an important role in the reduction of injuries and the protection of the player [3]. Given that ice hockey is a sport including a variety of complex movements, the main challenge in equipment design is to find the optimum protection without restricting the player during complex maneuvers.

Wearable sensor technologies that measure motion in real-world scenarios offer objective tools for building a scientific understanding of the motion of ice hockey player across a wide array of movements. Proven sensors include 3D accelerometers/gyroscopes for the performance analysis of biomechanical variables in skating [4] or the instrumentation of badminton rackets for stroke detection [5].

For evaluation of the equipment’s influence on the range of motion of players during complex maneuvers, a study is performed using the XSENS MVN Awinda (Xsens Technologies B.V. Netherlands) system to evaluate the influence of the equipment. The system utilizes 17 wireless Inertial Measurement Units (IMU) which are fitted to the body. Besides accelerometer and gyroscope raw data, the system can calculate additional parameters such as the velocities and positions of the body segments and the corresponding joint angles [6]. During the study, four different drills (sprint, slap shot, linear crossover, and power turn) are performed by 20 ice hockey players under the two conditions with and without equipment.

One of the challenges in the analysis of the data to find a difference between the two conditions is a large number of variables and their change over time. The goal of this master thesis is to determine the variables over time that are most relevant to distinguish between a drill performed with and without equipment. Layer-wise relevance propagation (LRP) [7], has successfully been applied in the identification of gait characteristics between subjects in walking [8] and overground running [9]. Therefore, we propose to implement LRP Lto identify the most relevant variables in the differentiation in drill kinematics when wearing protective equipment versus no equipment. In this regard, additional neural networks based on different variables such as joint angles will be implemented for each movement of interest to distinguish between the two conditions. Finally, LRP will be performed to determine the contribution of each variable to the classification result.

References

[1] P. Federolf and B. Nigg, “Skating performance in ice hockey when using a flared skate blade design,” Cold Reg. Sci. Technol., vol. 70, pp. 12–18, Jan. 2012, doi: 10.1016/j.coldregions.2011.08.009.
[2] J. T. Worobets, J. C. Fairbairn, and D. J. Stefanyshyn, “The influence of shaft stiffness on potential energy and puck speed during wrist and slap shots in ice hockey,” Sports Eng., vol. 9, no. 4, pp. 191–200, Dec. 2006, doi: 10.1007/BF02866057.
[3] C. Asplund, S. Bettcher, and J. Borchers, “Facial protection and head injuries in ice hockey: a systematic review,” Br. J. Sports Med., vol. 43, no. 13, pp. 993–999, Dec. 2009, doi: 10.1136/bjsm.2009.060152.
[4] B. J. Stetter, E. Buckeridge, S. R. Nigg, S. Sell, and T. Stein, “Towards a wearable monitoring tool for in-field ice hockey skating performance analysis,” Eur. J. Sport Sci., vol. 19, no. 7, pp. 893–901, Aug. 2019, doi: 10.1080/17461391.2018.1563634.
[5] T. Steels, B. Van Herbruggen, J. Fontaine, T. De Pessemier, D. Plets, and E. De Poorter, “Badminton Activity Recognition Using Accelerometer Data,” Sensors, vol. 20, no. 17, Art. no. 17, Jan. 2020, doi: 10.3390/s20174685.
[6] Xsens, “MVN Analyze.” https://www.xsens.com/products/mvn-analyze
[7] S. Bach, A. Binder, G. Montavon, F. Klauschen, K.-R. Müller, and W. Samek, “On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation,” PLOS ONE, vol. 10, no. 7, p. e0130140, Jul. 2015, doi: 10.1371/journal.pone.0130140.
[8] F. Horst, S. Lapuschkin, W. Samek, K.-R. Müller, and W. I. Schöllhorn, “Explaining the unique nature of individual gait patterns with deep learning,” Sci. Rep., vol. 9, no. 1, Art. no. 1, Feb. 2019, doi: 10.1038/s41598-019-38748-8.
[9] F. Hoitz, V. von Tscharner, J. Baltich, and B. M. Nigg, “Individuality decoded by running patterns: Movement characteristics that determine the uniqueness of human running,” PLOS ONE, vol. 16, no. 4, p. e0249657, Apr. 2021, doi: 10.1371/journal.pone.0249657.