Abstract

     Recent advances in wearable sensing and machine learning have created ample opportunities for “in the wild” movement analysis in sports, since the combination of both enables real-time feedback to be provided to athletes and coaches, as well as long-term monitoring of movements. The potential for real-time feedback is useful for performance enhancement or technique analysis, and can be achieved by training efficient models and implementing them on dedicated hardware. Long-term monitoring of movement can be used for injury prevention, among others. Such applications are often enabled by training a machine learned model from large datasets that have been collected using wearable sensors. Therefore, in this perspective paper, we provide an overview of approaches for studies that aim to analyze sports movement “in the wild” using wearable sensors and machine learning. First, we discuss how a measurement protocol can be set up by answering six questions. Then, we discuss the benefits and pitfalls and provide recommendations for effective training of machine learning models from movement data, focusing on data pre-processing, feature calculation, and model selection and tuning. Finally, we highlight two application domains where “in the wild” data recording was combined with machine learning for injury prevention and technique analysis, respectively.

Introduction

     Measuring sports movement during training and competition allows monitoring athletes’ performance and their risk of injury (Camomilla et al., 2018, Cust et al., 2019). Performance monitoring is relevant to assess motor capacity and physical demand, as well as to analyze technique and how technique impacts performance (Camomilla et al., 2018). In the complementary perspective of sport-related injuries, monitoring can be oriented to preventing, assessing, and informing the recovery from injuries (Preatoni et al., 2022).

     The state-of-the-art approach to measure movement is to perform a biomechanical analysis from data recorded in a lab environment using optical motion capture (OMC) systems and force plates. Such an analysis includes calculating spatio-temporal variables, as well as joint angles, joint moments, and ground reaction forces. Musculoskeletal models can provide additional insights into muscle forces and activations, which cannot be measured directly. However, these state-of-the-art measurement techniques can only be used in a laboratory environment and are limited by high costs, a stationary setup, and a short duration.

Conclusion

     In this perspective paper, we have discussed different opportunities and challenges regarding analysis of “in the wild” movement data using machine learning. We have highlighted the importance of careful data recordings. This is important when recording movement in general, to collect the correct data at the correct point in time, but becomes even more important when machine learning is used for processing, to ensure that the trained model works for the data it should be analyzing, and that the correct analysis is performed. We have discussed several considerations regarding both data recording in general and specific to training of machine learning models, to ensure that in future researchers can properly address these in experiments. Furthermore, we highlighted how “in the wild” movement data can be used in two application domains, specifically monitoring injury risk and technique analysis.

     By combining machine learning with “in the wild” recordings, the main advantages are the potential for real-time feedback, while analyses can also be performed on large datasets. The real-time potential can be used to enhance the performance of athletes during training and competition. Its advantages for large datasets can be exploited when performing long-term monitoring to forecast injuries, especially by developing personalized models, while statistical features could be found that provide more insight than traditional parameters as well. The combination of real-time feedback with analysis on large datasets allows for computationally efficient calculation of many variables, e.g. joint angles of a kinematic model, which can then be used to create avatars of athlete’s movements to provide actionable insights. In conclusion, this perspective paper provides researchers with guidance and directions for future development of machine learning models for movement analysis “in the wild”.