summary: Researchers have developed an AI model that can identify the emotional state of tennis players with high accuracy by analyzing their body language during matches. Trained on real-life footage, the AI can detect positive and negative emotions, but is better at recognizing negative ones. The technology could have applications in sports training, healthcare and other fields, but it has also raised ethical concerns about privacy and data misuse.
Key Facts:
- The AI model accurately identifies the emotions of tennis players based on their body language.
- Both AI and humans are good at recognizing negative emotions.
- There are potential applications in sports, healthcare and more, but ethical concerns need to be addressed.
sauce: kit
For the study “Recognizing Emotional States from Expressive Behavior of Tennis Players Using Convolutional Neural Networks”, researchers from the Department of Sports Science, Software Development and Computer Science at KIT and the University of Duisburg-Essen developed a special AI model.
They used a pattern recognition program to analyze video of tennis players recorded during actual matches.
Success rate: 68.9 percent
“Our model can identify emotional states with an accuracy of up to 68.9 percent, which is comparable to and sometimes even surpasses assessments by human observers and previous automated methods,” says Professor Darko Jekaucu from KIT's Institute of Sport and Sports Science.
A key and unique feature of this research is that the project team used real-life scenarios, rather than simulations or fictionalized scenarios, to train the AI system. The researchers recorded video sequences of 15 tennis players in specific situations, focusing on the body language they displayed when winning and losing points.
The video showed players giving cues such as bowing their heads, raising their arms to express joy, dangling their rackets, and varying walking speeds that can be used to identify the players' emotional states.
After feeding it this data, the AI learned to associate body language signals with different emotional responses and determine whether points were gained (positive body language) or lost (negative body language).
“Training in natural situations is a major advancement for identifying real emotional states, allowing predictions in real-life scenarios,” Jecauuk said.
Humans and machines recognize negative emotions better than positive ones
Not only does this study suggest that AI algorithms may one day surpass human observers in their ability to identify emotions, it also uncovers an even more interesting aspect: humans and AI are both good at recognizing negative emotions.
“This may be because negative emotions are expressed in a more overt way and therefore easier to identify,” Jekauk said.
“Psychological theories suggest that humans have evolved to be better able to recognize negative emotional expressions because quickly resolving conflict situations, for example, is essential for social cohesion.”
Ethical aspects need to be clarified before use
The research envisions applications of reliable emotion recognition across a range of sports, including improving training methods, team dynamics and performance, and preventing burnout. Other fields, such as healthcare, education, customer service, and automotive safety, could also benefit from reliable early detection of emotional states.
“While this technology has the potential to bring great benefits, we must also consider the potential risks that come with it, particularly those related to privacy and data misuse,” Jecauuk said.
“Our research strictly follows existing ethical guidelines and data protection regulations. And when considering practical applications of such techniques, it is essential to clarify ethical and legal issues in advance.”
About this news about AI and emotion research
author: Margarete Lene
sauce: kit
contact: Margarete Lene – KIT
image: Image courtesy of Neuroscience News
Original Research: Open access.
Justus Hartlieb et al., “Recognizing Emotional States from Tennis Players' Expressive Behavior Using Convolutional Neural Networks.” Knowledge-Based Systems
Abstract
Recognizing Emotional States of Tennis Players from Their Expressive Behaviors Using Convolutional Neural Networks
In this study, we present an AI model that leverages advanced Convolutional Neural Networks (CNNs) to recognize emotional states in real-world sports environments, specifically tennis matches.
In contrast to previous studies, which mainly used data obtained from actors and rudimentary statistical methods, this study focuses on analyzing bodily expressions in real-life situations, aiming to express human emotions more naturally.
Our CNN-based models show accuracy rates up to 68.9%, often outperforming or matching human observers. Interestingly, both machine learning models and human observers showed a common tendency to identify negative emotional states more effectively, which can be attributed to the expressions of these states being more intense and direct.
These results not only advance the state of the art in emotional state recognition, but also pave the way for broader applications including in healthcare and automotive safety, marking a major step forward in the development of sophisticated and universally applicable emotion recognition systems.
