Automatic Classification of Skating Cross-Country Skiing Sub- Techniques based on a Single Wearable Sensor and Biomechanical Models

Abstract

The aim of this study was to design and validate a sub-technique classification algorithm for everyday cross-country trainings, based on a single sensor worn on the upper back and biomechanical models. The sensor (FieldWiz, Advanced Sport Instruments, Switzerland) recorded GNSS data, acceleration, and angular velocity. Using a customized fusion algorithm and a trunk model, the athlete’s centre of mass kinematics were obtained. Cycles were detected based on maxima in trunk inclination and a Gaussian mixture model was used to assign each cycle to its corresponding sub-technique (Gear 2, 3, 4) based on cycle distance, amount of lateral excursion and trunk inclination periodicity. Gaussian mixture parameters were determined from a separate dataset of short roller skiing trials. The algorithm was validated against video recordings with 5 junior level athletes skating on a 2.4 km lap with roller skis at medium intensity. Turns were removed and uphill and flat sections were selected. 925 sec of data remained and each second was attributed one sub-technique. Gears 2, 3, 4 were skied during 81, 600, 244 sec, respectively. 98.4% of all seconds were correctly classified and misclassifications mainly happened during transitions. This approach of model-based sub-technique classification proved extremely efficient, is fully automatic, and can be used during daily trainings. On-snow validity should be assessed in the future and other sub-techniques (e.g. double poling) could be added.

Full text

German Journal of Exercise and Sport Research
Abstracts
the relationship between contextualized player motion (e. g. speed when
defending within own half) and technical-tactical performances during
competition.
Heart rate variability guided endurance training in recreational
runners
Christoph Zinner1, Daniela Schäfer Olstad2, Billy Sperlich3
1University of Applied Sciences for Police and Administration of Hesse,
Wiesbaden, Germany; 2Polar Electro Oy, Kempele, Finland; 3University of
Würzburg, Würzburg, Germany
e aim was to investigate whether heart rate variability (HRV) guided exer-
cise prescription yields comparable results on 5000 m running performance
and key components of endurance performance in recreational runners.
irty-one recreational runners were systematically parallelized to one of
two groups performing a 4-wk mesocycle with similar training intensity
distribution (100%TRIMP) followed by a 3-wk mesocycle with 50% in-
creased TRIMP compared to the rst 4-wk mesocycle, and one-wk taper-
ing. Both groups used similar individualized training plans with the HRV
group having their training adjusted based on a 6-minute HRV test by Po-
lar Electro Oy each morning during the second mesocycle. VO2peak and
running economy were assessed at baseline (T0), aer four (T1), seven
(T2), and eight weeks (T3).
HRV trained less sessions and with a lower mean intensity as CONTROL.
e 5000 m time decreased in CONTROL from T0 to T2 and T3, and from
T0 to T3 and T1 to T3 in HRV. VO2peak increased from T1 to T2 (p = 0.02)
with HRV and from T0 to T3 (p = 0.006) with control. Running economy
improved only from T0 to T3 and from T2 to T3 (p < 0.01) with HRV. An
individual mean response analysis indicated a high number of responders
(n = 8 of 16) in CON and in HRV (n = 9/13).
Despite less training time HRV guided training showed comparable im-
provements in 5000 m running performance. HRV guided training may be
a potential method to adjust exercise intensity and improve performance
in recreational runners.
Automatic Classification of Skating Cross-Country Skiing Sub-
Techniques based on a Single Wearable Sensor and Biomechanical
Models
Benedikt Fasel1, Matej Supej2, Marko Laaksonen3
1Archinisis GmbH, Fribourg, Switzerland; 2University of Ljubljana, Ljubljana,
Slovenia; 3Mid Sweden University, Östersund, Sweden
e aim of this study was to design and validate a sub-technique classi-
cation algorithm for everyday cross-country trainings, based on a sin-
gle sensor worn on the upper back and biomechanical models. e sen-
sor (FieldWiz, Advanced Sport Instruments, Switzerland) recorded GNSS
data, acceleration, and angular velocity. Using a customized fusion algo-
rithm and a trunk model [1], the athlete’s center of mass kinematics were
obtained. Cycles were detected based on maxima in trunk inclination and
a Gaussian mixture model was used to assign each cycle to its correspond-
ing sub-technique (Gear 2, 3, 4) based on cycle distance, amount of lateral
excursion and trunk inclination periodicity. Gaussian mixture parameters
were determined from a separate dataset of short roller skiing trials. e
algorithm was validated against video recordings with 5 junior level ath-
letes skating on a 2.4 km lap with roller skis at medium intensity. Turns
were removed and uphill and at sections were selected. 925 sec of data
remained, and each second was attributed one sub-technique. Gears 2, 3,
4 were skied during81, 600, 244 s, respectively. 98.4% of all seconds were
correctly classied and misclassications mainly happened during transi-
tions. is approach of model-based sub-technique classication proved
extremely ecient, is fully automatic, and can be used during daily train-
ings. On-snow validity should be assessed in the future and other sub-tech-
niques (e. g. double poling) could be added.
References
1. Fasel, et al. (2016). Remote Sensing. , 8(671) https://doi.org/10.3390/rs8080671.
mands of the intensity of resistance training, a category-ratio scale (CR10)
was used by the subjects aer each training session. e participants of
both groups trained twice a week for 9 weeks. e HAT and WUP pro-
grams used the same exercises, the same total training volume and the
same total intensity in these six weeks. e dierence between the two
programs was in the distribution within each training phase. e HAT and
WUP groups trained using a periodized strength programs with all pro-
grams variables controlled (e. g., volume and intensity). e HAT group
used a linear not varying intensity, whereas the WUP group had a varied
intensity. e results show that both the HAT and WUP groups made sig-
nicant (p ≤ 0.05) increases in strength and power. us, HAT and WUP
are similarly eective over a nine-week training period, and the decision
to use HAT or WUP depends on the preferences of the individual athlete.
[YIA] Periodization of plyometrics: is there an optimal overload
principle?
Maarten Lievens, Jan Bourgois, Jan Boone
Ghent University, Ghent, Belgium
is study investigated the acute and chronic eects of three plyomet-
ric training (PT) programs with equal training loads (intensity ×
volume
× frequency) on speed, agility and jumping performance. Forty-four male
recreational team sport athletes were either assigned to a program that
(1) increased training volume with exercises of mixed intensity (Mix), (2)
kept training volume equal and increased exercise intensity (LowHi), (3)
increased training volume and kept exercise intensity low (Low) or to a
(4) control group (Control). Subjects trained twice a week for 8 weeks and
were tested for 5 m (5 m) and 10 m sprint (10 m), 5 × 10 m shuttle run (5
× 10 m), squat jump (SJ), countermovement jump without (CMJ) and with
arm swing (CMJa) and standing broad jump (SBJ). e change in 5 m,
10 m, 5 × 10 m and SJ performance did not signicantly (p > 0.05) dier
between groups. Sprinting and agility did not change aer 8 weeks of PT
(p > 0.05). e CMJ, CMJa and SBJ increased in the PT groups compared
to the control group (p < 0.05). ere was no dierence (p > 0.05) between
PT groups. Additionally, it was shown that a training session of high in-
tensity was more likely to diminish performance the following days. To
conclude, PT programs following a dierent overload pattern, i.
e. dier-
ent combination of volume and intensity, but equal training load showed
similar performance eects in recreationally trained men. However, prior
to competition, a PT of low intensity is preferred over a PT of high inten-
sity in order to avoid a decline in performance.
Monitoring with Wearable Technology
The use of higher dimensional analyses to visualize the training
process
Dan Weaving
Leeds Rhinos Rugby League Club, Leeds Beckett University, Leeds, United
Kingdom
Quantifying the training load imposed onto team sports athletes is com-
plex given the concurrent multi-modal training programs that these ath-
letes undertake. Consequently, in the age of technology, a wealth of data
representing dierent aspects of the training process are collected. To pre-
vent data overload, optimizing how this data travels from collection to
presentation to coaches is crucial to embed data into decision making. Us-
ing training and competition load data collected over three seasons, this
presentation will provide an overview of how we have embedded high-
er dimensional analyses in professional rugby league practice to visualize
and communicate the relationship between multiple variables relating to
the training process and its outcomes (e. g. injury, performance). In par-
ticular, the use of principal component analysis to visualize the dierences
in external and internal training intensities of technical-tactical training
drills. Also, the use of partial least squares correlation analysis to visualize