May, 09th, 2024. In a model experiment, a research team from KIT Karlsruhe and the University of Duisburg Essen has succeeded in identifying emotional states in athletes by means of AI observation. A more detailed evaluation revealed that both AI and human observation skills are better at recognizing negative emotions.

In the study „Recognition of emotional states from the expressive behavior of tennis players using convolutional neural networks“, a research team led by study leader Professor Darko Jekauc from the Institute of Sport and Sports Science (IfSS) at the Karlsruhe Institute of Technology and the University of Duisburg Essen was able to develop an AI model. This involved recording and analyzing fixed video sequences during real scenes of tennis players during real competitions.

The focus of both the observation and the AI conditioning was on body language when winning and losing points. Cues such as a lowered head, cheering arms thrown in the air, a hanging bat or differences in walking speed were learned and identified by pattern recognition.

„Our model recognizes affective states with an accuracy of up to 68.9 percent. This is comparable with the evaluation by human observers as well as with earlier automated methods, and in some cases even superior to them,“ says Professor Darko Jekauc from KIT Karlsruhe.

During the evaluation of the results, the research team discovered a further phenomenon: both AI and humans are better at recognizing negative emotions than positive emotions. Professor Jekauc explains this as „… that these are easier to identify due to their clearer forms of expression“ „Psychological theories suggest that humans are evolutionarily more attuned to perceiving negative emotional expressions. For example, because it is crucial for the social context to defuse conflicts in good time.“

„Although this technology offers significant advantages in the future, the potential risks associated with it must be taken into account, especially in terms of privacy and data misuse,“ emphasizes Jekauc. „Our study was strictly guided by existing ethical guidelines and data protection regulations. Also with a view to future applications of such technology in practice, it is essential to clarify ethical and legal issues in advance.“

Original publication

Darko Jekauc, Diana Burkart, Julian Fritsch, Marc Hesenius, Ole Meyer, Saquib Sarfraz, Rainer Stiefelhagen: Recognizing affective states from the expressive behavior of tennis players using convolutional neural networks. Knowledge-Based Systems, Vol. 295, 2024. DOI: 10.1016/j.knosys.2024.111856

Recognizing affective states from the expressive behavior of tennis players using convolutional neural networks

https://www.sciencedirect.com/science/article/pii/S0950705124004908?via%3Dihub