Tobias Baumgartner
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Tobias Baumgartner has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Stefanie Klatt
Institute of Exercise Training and Sport Informatics, German Sport University Cologne, Cologne, Germany
5 shared publications - Lars Donath
Department of Applied Exercise Science, German Sport University Cologne, Cologne, Germany
2 shared publications - André Nicklas
Institute of Exercise Training and Sport Informatics, German Sport University Cologne, Cologne, Germany
1 shared publication - Benjamin Paaßen
Faculty of Technology, Bielefeld University, Bielefeld, Germany
1 shared publication - Fabiola Haas
Department for Physical Activity in Public Health, Institute of Movement and Neurosciences, German Sport University Cologne, Am Sportpark Müngersdorf 6, 50933 Cologne, Germany
1 shared publication - Florian Seifriz
Department of Cognitive and Team/Racket Sport Research, Institute of Exercise Training and Sport Informatics, German Sport University Cologne, 50933 Cologne, Germany
1 shared publication
Latest Publications
- 6 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 17, 642-652
Extended Reality as a Training Approach for Visual Real-Time Feedback in Golf
Visual feedback can enhance motor learning in golf, however, so far, this has mostly been time-delayed. Using Extended Reality (XR), learners can receive visual feedback in real-time without relying on an external perspective. The aim of this study was to investigate a novel XR-based method of real-time feedback for golf putt training. Thirty-two subjects were divided between a XR-group and a technique instruction (TI) training group, completing three weeks of training each, and a no-training control (NTc) group. XR training included a virtual club projected into the real training environment via HoloLens 2. Real-time information was provided on the pre-calculated targeted club alignment and velocity. The TI method included written technical instructions already used in golf putt training. Before and after the training, all subjects performed a pre-, post- and retention-test, and two transfer-tests with shorter (2 m) and longer (4 m) putting distances than during training (3 m). Results showed a significant increase in the number of hits and a significant decrease in the radial error after training in all groups. The number of hits remained the same during the retention-test and at shorter putting distance, but decreased at longer distance. The radial error decreased even more during the retention-test and revealed no difference between putting distances. Our study shows that the acquisition of golf putting skills can be enhanced not only through a proven technique instruction training, but also by incorporating visual real-time feedback via XR. Future studies could examine the extent to which the effects of this method exceed those of technique instruction training over a longer period when learners can familiarize themselves with it more.
- 6 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 17, 642-652
Extended Reality as a Training Approach for Visual Real-Time Feedback in Golf
Visual feedback can enhance motor learning in golf, however, so far, this has mostly been time-delayed. Using Extended Reality (XR), learners can receive visual feedback in real-time without relying on an external perspective. The aim of this study was to investigate a novel XR-based method of real-time feedback for golf putt training. Thirty-two subjects were divided between a XR-group and a technique instruction (TI) training group, completing three weeks of training each, and a no-training control (NTc) group. XR training included a virtual club projected into the real training environment via HoloLens 2. Real-time information was provided on the pre-calculated targeted club alignment and velocity. The TI method included written technical instructions already used in golf putt training. Before and after the training, all subjects performed a pre-, post- and retention-test, and two transfer-tests with shorter (2 m) and longer (4 m) putting distances than during training (3 m). Results showed a significant increase in the number of hits and a significant decrease in the radial error after training in all groups. The number of hits remained the same during the retention-test and at shorter putting distance, but decreased at longer distance. The radial error decreased even more during the retention-test and revealed no difference between putting distances. Our study shows that the acquisition of golf putting skills can be enhanced not only through a proven technique instruction training, but also by incorporating visual real-time feedback via XR. Future studies could examine the extent to which the effects of this method exceed those of technique instruction training over a longer period when learners can familiarize themselves with it more.
- Open Access
- 28 August 2023
- Published by Springer Nature in Scientific Reports
- Vol. 13 (1) , 1-11
Extracting spatial knowledge from track and field broadcasts for monocular 3D human pose estimation
Collecting large datasets for investigations into human locomotion is an expensive and labor-intensive process. Methods for 3D human pose estimation in the wild are becoming increasingly accurate and could soon be sufficient to assist with the collection of datasets for analysis into running kinematics from TV broadcast data. In the domain of biomechanical research, small differences in 3D angles play an important role. More precisely, the error margins of the data collection process need to be smaller than the expected variation between athletes. In this work, we propose a method to infer the global geometry of track and field stadium recordings using lane demarcations. By projecting estimated 3D skeletons back into the image using this global geometry, we show that current state-of-the-art 3D human pose estimation methods are not (yet) accurate enough to be used in kinematics research.
- 28 August 2023
- Published by Springer Nature in Scientific Reports
- Vol. 13 (1) , 1-11
Open AccessExtracting spatial knowledge from track and field broadcasts for monocular 3D human pose estimation
Collecting large datasets for investigations into human locomotion is an expensive and labor-intensive process. Methods for 3D human pose estimation in the wild are becoming increasingly accurate and could soon be sufficient to assist with the collection of datasets for analysis into running kinematics from TV broadcast data. In the domain of biomechanical research, small differences in 3D angles play an important role. More precisely, the error margins of the data collection process need to be smaller than the expected variation between athletes. In this work, we propose a method to infer the global geometry of track and field stadium recordings using lane demarcations. By projecting estimated 3D skeletons back into the image using this global geometry, we show that current state-of-the-art 3D human pose estimation methods are not (yet) accurate enough to be used in kinematics research.
- Open Access
Monocular 3D Human Pose Estimation for Sports Broadcasts using Partial Sports Field Registration
The filming of sporting events projects and flattens the movement of athletes in the world onto a 2D broadcast image. The pixel locations of joints in these images can be detected with high validity. Recovering the actual 3D movement of the limbs (kinematics) of the athletes requires lifting these 2D pixel locations back into a third dimension, implying a certain scene geometry. The well-known line markings of sports fields allow for the calibration of the camera and for determining the actual geometry of the scene. Close-up shots of athletes are required to extract detailed kinematics, which in turn obfuscates the pertinent field markers for camera calibration. We suggest partial sports field registration, which determines a set of scene-consistent camera calibrations up to a single degree of freedom. Through joint optimization of 3D pose estimation and camera calibration, we demonstrate the successful extraction of 3D running kinematics on a 400m track. In this work, we combine advances in 2D human pose estimation and camera calibration via partial sports field registration to demonstrate an avenue for collecting valid large-scale kinematic datasets. We generate a synthetic dataset of more than 10k images in Unreal Engine 5 with different viewpoints, running styles, and body types, to show the limitations of existing monocular 3D HPE methods. Synthetic data and code are available at https://github.com/tobibaum/PartialSportsFieldReg_3DHPE.
Open AccessMonocular 3D Human Pose Estimation for Sports Broadcasts using Partial Sports Field Registration
The filming of sporting events projects and flattens the movement of athletes in the world onto a 2D broadcast image. The pixel locations of joints in these images can be detected with high validity. Recovering the actual 3D movement of the limbs (kinematics) of the athletes requires lifting these 2D pixel locations back into a third dimension, implying a certain scene geometry. The well-known line markings of sports fields allow for the calibration of the camera and for determining the actual geometry of the scene. Close-up shots of athletes are required to extract detailed kinematics, which in turn obfuscates the pertinent field markers for camera calibration. We suggest partial sports field registration, which determines a set of scene-consistent camera calibrations up to a single degree of freedom. Through joint optimization of 3D pose estimation and camera calibration, we demonstrate the successful extraction of 3D running kinematics on a 400m track. In this work, we combine advances in 2D human pose estimation and camera calibration via partial sports field registration to demonstrate an avenue for collecting valid large-scale kinematic datasets. We generate a synthetic dataset of more than 10k images in Unreal Engine 5 with different viewpoints, running styles, and body types, to show the limitations of existing monocular 3D HPE methods. Synthetic data and code are available at https://github.com/tobibaum/PartialSportsFieldReg_3DHPE.
- Open Access
Revealing the Mutual Information between Body-Worn Sensors and Metabolic Cost in Running
Running power is a popular measure to gauge objective intensity. It has recently been shown, though, that foot-worn sensors alone cannot reflect variations in the exerted energy that stems from changes in the running economy. In order to support long-term improvement in running, these changes need to be taken into account. We propose leveraging the presence of two additional sensors worn by the most ambitious recreational runners for improved measurement: a watch and a heart rate chest strap. Using these accelerometers, which are already present and distributed over the athlete’s body, carries more information about metabolic demand than a single foot-worn sensor. In this work, we demonstrate the mutual information between acceleration data and the metabolic demand of running by leveraging the information bottleneck of a constrained convolutional neural network. We perform lab measurements on 29 ambitious recreational runners (age = 28 ± 7 years, weekly running distance = 50 ± 25 km, = 60.3 ± 7.4 mL · min−1·kg−1). We show that information about the metabolic demand of running is contained in kinetic data. Additionally, we prove that the combination of three sensors (foot, torso, and lower arm) carries significantly more information than a single foot-worn sensor. We advocate for the development of running power systems that incorporate the sensors in watches and chest straps to improve the validity of running power and, thereby, long-term training planning.Open AccessRevealing the Mutual Information between Body-Worn Sensors and Metabolic Cost in Running
Running power is a popular measure to gauge objective intensity. It has recently been shown, though, that foot-worn sensors alone cannot reflect variations in the exerted energy that stems from changes in the running economy. In order to support long-term improvement in running, these changes need to be taken into account. We propose leveraging the presence of two additional sensors worn by the most ambitious recreational runners for improved measurement: a watch and a heart rate chest strap. Using these accelerometers, which are already present and distributed over the athlete’s body, carries more information about metabolic demand than a single foot-worn sensor. In this work, we demonstrate the mutual information between acceleration data and the metabolic demand of running by leveraging the information bottleneck of a constrained convolutional neural network. We perform lab measurements on 29 ambitious recreational runners (age = 28 ± 7 years, weekly running distance = 50 ± 25 km, = 60.3 ± 7.4 mL · min−1·kg−1). We show that information about the metabolic demand of running is contained in kinetic data. Additionally, we prove that the combination of three sensors (foot, torso, and lower arm) carries significantly more information than a single foot-worn sensor. We advocate for the development of running power systems that incorporate the sensors in watches and chest straps to improve the validity of running power and, thereby, long-term training planning. - Open Access
- 14 January 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (2) , 1139
Heatmap Analysis to Differentiate Diverse Player Types in Table Tennis—A Training and Tactical Strategy Development Potential
(1) Background: Computer-based analyses have been widely used to study aspects of various team and racket sports. However, such analyses have so far eluded table tennis, except in special competition forms, under fixed laboratory conditions or have only tracked the ball. The aim was to detect a basic, global, positional behavior, independent of the score. (2) Methods: We investigated the player position of professional male table tennis players with respect to handedness (right/left), playing system (offensive/defensive) and racket holding (shakehand/penholder) to determine the applicability of automated analysis systems. We used existing video data of competitive matches (N = 198 sets; 2006–2020) and transliterated them into an x–y coordinate system. From this, we were able to conduct a heatmap analysis for different types of players. (3) Results: The comparison between right- and left-handed players resulted in a significant difference in the positioning of the x coordinate (D = 0.5663; p = 0.001). Both groups positioned themselves on average in their own backhanded half of the table (Re: x = −0.22 m, Li: x = 0.39 m). (4) Conclusions: Our results have yielded valuable insights into the importance of analyzing positional behavior in a differentiated manner depending on handedness, playing strategy and racket holding posture.
- 14 January 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (2) , 1139
Open AccessHeatmap Analysis to Differentiate Diverse Player Types in Table Tennis—A Training and Tactical Strategy Development Potential
(1) Background: Computer-based analyses have been widely used to study aspects of various team and racket sports. However, such analyses have so far eluded table tennis, except in special competition forms, under fixed laboratory conditions or have only tracked the ball. The aim was to detect a basic, global, positional behavior, independent of the score. (2) Methods: We investigated the player position of professional male table tennis players with respect to handedness (right/left), playing system (offensive/defensive) and racket holding (shakehand/penholder) to determine the applicability of automated analysis systems. We used existing video data of competitive matches (N = 198 sets; 2006–2020) and transliterated them into an x–y coordinate system. From this, we were able to conduct a heatmap analysis for different types of players. (3) Results: The comparison between right- and left-handed players resulted in a significant difference in the positioning of the x coordinate (D = 0.5663; p = 0.001). Both groups positioned themselves on average in their own backhanded half of the table (Re: x = −0.22 m, Li: x = 0.39 m). (4) Conclusions: Our results have yielded valuable insights into the importance of analyzing positional behavior in a differentiated manner depending on handedness, playing strategy and racket holding posture.
- Open Access
Limitations of Foot-Worn Sensors for Assessing Running Power
Running power as measured by foot-worn sensors is considered to be associated with the metabolic cost of running. In this study, we show that running economy needs to be taken into account when deriving metabolic cost from accelerometer data. We administered an experiment in which 32 experienced participants (age = 28 ± 7 years, weekly running distance = 51 ± 24 km) ran at a constant speed with modified spatiotemporal gait characteristics (stride length, ground contact time, use of arms). We recorded both their metabolic costs of transportation, as well as running power, as measured by a Stryd sensor. Purposely varying the running style impacts the running economy and leads to significant differences in the metabolic cost of running (p < 0.01). At the same time, the expected rise in running power does not follow this change, and there is a significant difference in the relation between metabolic cost and power (p < 0.001). These results stand in contrast to the previously reported link between metabolic and mechanical running characteristics estimated by foot-worn sensors. This casts doubt on the feasibility of measuring running power in the field, as well as using it as a training signal.
Open AccessLimitations of Foot-Worn Sensors for Assessing Running Power
Running power as measured by foot-worn sensors is considered to be associated with the metabolic cost of running. In this study, we show that running economy needs to be taken into account when deriving metabolic cost from accelerometer data. We administered an experiment in which 32 experienced participants (age = 28 ± 7 years, weekly running distance = 51 ± 24 km) ran at a constant speed with modified spatiotemporal gait characteristics (stride length, ground contact time, use of arms). We recorded both their metabolic costs of transportation, as well as running power, as measured by a Stryd sensor. Purposely varying the running style impacts the running economy and leads to significant differences in the metabolic cost of running (p < 0.01). At the same time, the expected rise in running power does not follow this change, and there is a significant difference in the relation between metabolic cost and power (p < 0.001). These results stand in contrast to the previously reported link between metabolic and mechanical running characteristics estimated by foot-worn sensors. This casts doubt on the feasibility of measuring running power in the field, as well as using it as a training signal.