The research area “Health Tech” at MCI includes research on the application of technology to medicine, sports and tourism. Bernhard Hollaus gave a talk at the ECSS-Congress 2021, (European College of Sport Science) a professional conference ranking among the two best in its field worldwide.
His research is about the recognition of different swing types in tennis. Since in tennis and other sports, data about the opponent and about oneself is becoming increasingly important for success. But the data alone are of no use: Only the correct interpretation can lead to success. This is exactly where the methods of "machine learning" come in because they enable the calculation of "expected goals" and "targets", but also the swing recognition in tennis.
In tennis in particular, it is interesting to see which stroke types are performed most often by the athletes, which stroke sequences lead to the highest score, and how the athletes' stroke power decreases with fatigue. The challenge of data acquisition lies in the autonomous measurement of a stroke. A suitable sensor and algorithm are needed for this machine recognition and evaluation of the techniques.For this, a "motion sensor" was used, which provides rotation rates, accelerations and magnetic flux densities. In this way, over 5,000 data sets were collected in an experiment with many tennis players. These data sets were then used for the training of a neural network to automatically assign a data set to a stroke type. Through this procedure, an accuracy of 94% could be achieved. That means that only 6% of the data sets were misclassified. In tennis, it is therefore already possible to use "machine learning" to provide athletes with relevant data and to analyze and optimize their opponent's playing style and their own.
For further information, please get in touch with:
Bernhard HollausSenior LecturerMedical, Health and Sports Technologies+43 512 firstname.lastname@example.org
ECSS Congress 2021 – Bernhard Hollaus giving his talk on “Tennis Shot Classification using a wearable and neural networks”. © MCI-Hollaus.