Technique analysis in elite athletes using principal component analysis

Abstract

The aim of this study was to advance current movement analysis methodology to enable a technique analysis in sports facilitating (1) concurrent comparison of the techniques between several athletes; (2) identification of potentially beneficial technique modifications and (3) a visual representation of the findings for feedback to the athletes. Six elite cross-country skiers, three world cup winners and three national elite, roller ski skated using the V2 technique on a treadmill while their movement patterns were recorded using 41 reflective markers. A principal component analysis performed on the marker positions resulted in multi-segmental "principal" movement components (PMs). A novel normalisation facilitated comparability of the PMs between athletes. Additionally, centre of mass (COM) trajectories were modelled. We found correlations between the athletes' performance levels (judged from race points) and specific features in the PMs and in the COM trajectories. Plausible links between COM trajectories and PMs were observed, suggesting that better performing skiers exhibited a different, possibly more efficient use of their body mass for propulsion. The analysis presented in the current study revealed specific technique features that appeared to relate to the skiers' performance levels. How changing these features would affect an individual athlete's technique was visualised with animated stick figures.

Should the Minimal Intervention Principle Be Considered When Investigating Dual-Tasking Effects on Postural Control?

“…All of the following PCA-based calculations were computed using a software package called "PManalyzer" [39], which is provided through open access. As a first analysis step, gaps in the marker trajectories were filled [40,41], then the data from each trial was normalized by subtracting the mean posture and dividing by the trial's mean Euclidean distance [29,32], finally, the marker coordinates were weighed according to the relative body mass, which they represent [29,42]. This normalization was designed to remove anthropometric differences while preserving the relative amplitude of the marker movements and to ensure that each participant equally affected the PCA output [29,32].…”

Machine and deep learning for sport-specific movement recognition: a systematic review of model development and performance

“…These identified variables 143 are represented in a compact feature variable (Mannini & Sabatini, 2010). (Gløersen, Myklebust, Hallén, & Federolf, 2018;Young & 145 Reinkensmeyer, 2014), vector coding techniques (Hafer & Boyer, 2017) and empirical cumulative 146 distribution functions (ECDF) (Plötz, Hammerla, & Olivier, 2011). An ECDF approach has been 147 shown to be advantageous over PCA as it derives representations of raw input independent of the and analysis of IMU data for sports application and vision-based human activity recognition, see and (Bux et al, 2017), respectively.…”

Real-time feedback by wearables in running: Current approaches, challenges and suggestions for improvements

“…(Tenforde, Borgstrom, Outerleys, & Davis, 2019)) and previous running experience. Indeed, several studies have shown technique to differ between (Brisson & Alain, 1996;Glazier & Lamb, 2017;Gloersen, Myklebust, Hallen, & Federolf, 2018;Morriss, Bartlett, & Fowler, 1997) and within individuals (Glazier & Lamb, 2017;Horst, Eekhoff, Newell, & Schollhorn, 2017;Riza, 2017). It can therefore be questioned to what extent an "ideal" technique should be aspired, for example, by using deviations from the average movement by 1 standard deviation as a criterion for technique modification (Bowser et al, 2018).…”