Techn. Fakultät Willkommen am Institut für Informatik FAU-Logo

Coaching and Feedback

Classification in Sports Events based on Orientation Features
  • The application of pattern recognition algorithm to sports can lead to innovative methods of sports analysis and visualization as well as to improvements in the performance of athletes. A classification of sports events can basically be applied to any game sports, reaching from classifying kicks in soccer to strokes in tennis or tricks in skateboarding. In this project, inertial sensors are attached to athletes or sport devices and the collected data are used for further analysis and classification.

    The basic classification pipeline contains a preprocessing step, feature extraction and the actual classification. Whereas previous projects already considered the extraction of generic features (e.g. statistical values of the inertial data), our goal is to define sports-related expert features. These expert features will based on the orientation of the sensor and thereby the orientation of the athlete or sport device.



    A typical example application is given by skateboarding. Most skateboard tricks show a typical signal sequence of the board’s rotation (see following figures). Although a classification that is only based on generic features already provides a reliable trick classification, future work will be to include information about the orientation of the board as expert features and thereby obtain even more significant results.

    Journal Articles
    Groh, Benjamin; Flaschka, Jasmin; Wirth, Markus; Kautz, Thomas; Fleckenstein, Martin; Eskofier, Björn
    Wearable Real-Time Skateboard Trick Visualization and Its Community Perception
    IEEE Computer Graphics and Applications, vol. 36, no. 5, pp. 12-18, 2016 (BiBTeX, Who cited this?)
    Articles in Conference Proceedings
    Groh, Benjamin; Fleckenstein, Martin; Eskofier, Björn
    IEEE EMBS 13th Annual International Body Sensor Networks Conference, San Francisco, USA, 15.06.2016, pp. 89-93, 2016 (BiBTeX, Who cited this?)
    - (KDD Workshop on Large-Scale Sports Analytics), Sydney, Australia, 10.08.2015, pp. 1-4, 2015 (BiBTeX, Who cited this?)
    Workshop on Large-Scale Sports Analytics (21st ACM SIGKDD Conference on Knowledge Discovery and Data Mining), Sydney, Australia, 10.08.2015, pp. 1-4, 2015 (BiBTeX, Who cited this?)
Putt Coach: Kinematic putt analysis for feedback training and machine learning
  • A mobile system for golf putt analysis and feedback is of high interest for technique training in the field and motor learning research. Recently, instrumented golf clubs comprising inertial measurement units (IMU) were introduced. However, a mobile analysis and augmented feedback system has not been realized so far.

    We developed such a system that featured automatic putt detection and parameter calculation as well as a closed feedback loop. With this system, data were collected to assess the motor learning effects of repetitive training. We analysed the training progress of eight players (2056 putts) based on 31 kinematic parameters.

    Our system detected more than 83 % of the putts for most of the participants while maintaining a false positive rate of 2.4 %. The training progress analysis revealed interesting insight of how relevant kinematic parameters change on a learning path. This feature is an enhancement to the classical pre-post analysis in motor learning research and can be used as to adjust training programs.

    In principle, the presented system can be used to realize mobile augmented feedback training for various sports disciplines beyond golf putting. It therefore provides an interesting tool for the analysis of motor learning processes in more detail.

    Articles in Conference Proceedings
    Jensen, Ulf; Dassler, Frank A.; Schmidt, Marcus; Hennig, Markus; Jaitner, Thomas; Eskofier, Björn
    Book of Abstracts of the 19th Annual Congress of the European College of Sport Science (19th Annual Congress of the European College of Sport Science), Amsterdam, The Netherlands, July 2-5, 2014, pp. 207-208, 2014, ISBN 978-94-622-8477-7 (BiBTeX, Who cited this?)
Drop Jump Ground Contact Time Measurement Based on 6-D Inertial Data
  • Augmented feedback of relevant training parameters is of great interest for athletes and coaches. This is especially the case if parameters cannot be visually determined and if small value deviations change the training effects. In plyometric training, the ground contact time was identified as crucial measures to monitor and optimize training. This project investigates the ground contact measurement of the drop jump with Inertial Measurement Unit measurements and a Hidden Markov Model (HMM) analysis. Different approaches to increase the jump phase detection performance like signal-shift boosting and more complex HMM modeling were investigated. Our algorithms achieved an absolute ground contact time error of 12.3 ms (signal-shift boosting) and 4.3 ms (complex HMM model) on a dataset of 8 subjects (80 jumps). The proposed measurement setup and analysis algorithm can be used to create a wearable augmented feedback system for plyometric training practice.

    Articles in Conference Proceedings
    Jensen, Ulf; Heinrich, Axel; Eskofier, Björn
    3D Analysis of Human Movement, 13 th International Symposium on (13th International Symposium on 3D Analysis of Human Movement (3D-AHM)), Lausanne, Switzerland, July 14 - July 17, 2014, pp. 311-314, 2014 (BiBTeX, Who cited this?)

    Signal-shift boosting: Analysis code and data download

    Complex HMM modeling: Analysis code and data download

Sensor-based Instant Golf Putt Feedback
  • This paper presents a golf putt feedback system based on sensor data collected at the putter head. The system comprises of out-of-the-box body area network components that are lightweight, mobile and inexpensive. More specifically, we employ a SHIMMER™ sensor node with an inbuilt three-axis gyroscope and accelerometer for measurement and an Android™ smartphone for data analysis and presentation.

    We collected data from five expert and six completely inexperienced subjects in order to facilitate the development of our algorithms. In a first step, we used a matched filter algorithm that enabled us to segment the unrestricted putting data into actual golf putts and other movements like training swings. In a second step, we analyzed the stroke phases in more depth in order to distinguish expert from inexperienced players. This was facilitated by extracting predefined movement parameters as a basis for explicit putt feedback.

    Finally, we transferred the algorithms that were developed during an offline data analysis to an Android implementation. It is capable of detecting putts and displaying instant feedback to guide players to a better putt execution. Therefore, the system can be used as an innovative personal putting coach for real-world use.

    Articles in Conference Proceedings
    Proceedings of the IACSS 2011 (9th International Symposium on Computer Science in Sport), Shanghai, P.R. China, September 21-24 2011, pp. 49-53, 2011, ISBN 978-1-84626-087-2 (BiBTeX, Who cited this?)
Athlete support systems

  • Inertial Sensor-Based Approach for Event Classification in Team Sports

    Coaches and players in team sports, e.g. soccer are heavily interested in statistics like number of shots and passes during training sessions. Currently available systems use video or computerized technology to create such statistics and are therefore mainly applicable for elite teams. Thus, the aim is to develop low-cost inertial sensor-based approaches for event classification, which can be used by teams at amateur level.


    Classification of Surfaces and Inclinations During Outdoor Running using Shoe-Mounted Inertial Sensors

    Embedded mobile systems for analysis and classification become more and more important in the field of sports and sports science.  Small and lightweight sensors in sportswear offer the possibility to monitor the athletes in a realistic environment, e.g.  during an outdoor run.  During the activity, the sportswear can automatically adapt to the current environment and hence optimizes the performance of the athlete. A major need is a running shoe, which can automatically be adapted to the current ground.

    Automatic GPS - based Labeling of Kinematic Data

    The main focus lies on the examination of kinematic data of outdoor conditions like grass or street. Therefore the kinematic data has to be labeled. The next figure shows the location of the runner in a certain outdoor condition.

    Automatic Step Segmentation

    An automatic step segmentation approach offers the possibility to examine kinematic data of single steps.

    Automatic Classification of Sports Exercises for Training Support

    An objective measurement is desired for monitoring parameters of athletes during workouts. Inertial sensors can be applied e.g. for the determination of different exercises in a training session. With such a system, the athlete does not have to write a self-report with the ordering of the performed exercises and can concentrate on the exercise itself. This has the potential to improve the sports performance of athletes.  

    Your Personal Movie Producer: Generating Highlight Videos in Soccer Using Wearables

    Nowadays, soccer videos are highly attractive internationally. Since the value of soccer videos drops after a while, highlights often appear to be more interesting than the whole video sequence. Manually browsing through long soccer videos, selecting highlight scenes, and applying video effects are time-consuming. Automatic approaches are currently based on high-quality TV broadcast material and are therefore mainly applicable for professional teams. The personal low-cost movie producer uses wearables for highlight video production. This movie producer based on wearables is a novel idea to provide amateur players with attractive automatically generated highlight videos.

    Articles in Conference Proceedings
    Sportinformatik 2012 : 9. Symposium der Sektion Sportinformatik der Deutschen Vereinigung für Sportwissenschaft (9. Symposium der Sektion Sportinformatik der Deutschen Vereinigung für Sportwissenschaft), Konstanz, Germany, 12.09.2012 - 14.09.2012, pp. 214-219, 2012 (BiBTeX, Who cited this?)
    Pattern Recognition (ICPR), 2012 21st International Conference on (21st International Conference on Pattern Recognition), Tsukuba, Japan, November 11-15, 2012 12, pp. 2258-2261, 2012 (BiBTeX, Who cited this?)