AI Can Identify the Emotions of Athletes
A novel emotion analysis model capable of identifying emotional states with accuracy comparable to humans.
Researchers at the Karlsruhe Institute of Technology (KIT) and the University of Duisburg-Essen have successfully utilized computer-assisted neural networks to accurately detect emotional states from the body language of tennis players during matches.
This study marks the first instance where they trained an artificial intelligence (AI) model using real-game data. Published in the journal Knowledge-Based Systems, their research demonstrates AI’s capability to interpret body language and emotions with a level of accuracy comparable to humans. However, it also raises ethical considerations.
Development of a Specialized AI Model for Emotion Recognition in Tennis Players
In their study titled “Recognizing affective states from the expressive behavior of tennis players using convolutional neural networks,” interdisciplinary teams from sports sciences, software development, and computer science at KIT and the University of Duisburg-Essen developed a specialized AI model. They employed pattern-recognition algorithms to analyze video footage of tennis players captured during live matches.
“Our model achieves an accuracy of up to 68.9 percent in identifying affective states, which compares favorably and sometimes exceeds assessments made by both human observers and earlier automated methods,” explained Professor Darko Jekauc from KIT’s Institute of Sports and Sports Science.
Utilization of Real-Life Scenarios in Training the AI System
A notable and distinctive aspect of the study is the team’s use of authentic, real-life scenarios rather than simulated or artificial settings to train their AI system. The researchers recorded video sequences featuring 15 tennis players in specific match conditions, focusing on their body language reactions when winning or losing points.
These videos captured players exhibiting cues such as lowered heads, raised arms in celebration, dangling rackets, or variations in walking speed, which helped discern their affective states.
By analyzing this dataset, the AI learned to correlate specific body language signals with various emotional reactions and accurately distinguish between positive (winning a point) and negative (losing a point) affective states. “Training in natural contexts represents a significant advancement for identifying genuine emotional states and enables predictions in real-world scenarios,” noted Jekauc.
The research not only suggests that AI algorithms may eventually outperform human observers in identifying emotions but also uncovered an intriguing finding: both humans and AI excel at recognizing negative emotions.
Recognition of Negative Emotions in Emotion Detection
“The reason for this could be that negative emotions are more straightforward to identify as they are expressed more visibly,” explained Jekauc. “Psychological theories propose that people are evolutionarily more adept at detecting negative emotional cues, possibly because swiftly resolving conflict situations is crucial for maintaining social harmony.”
The study envisions numerous sports applications for accurate emotion recognition, such as enhancing training methodologies, team dynamics, performance optimization, and mitigating burnout. Additionally, other fields like healthcare, education, customer service, and automotive safety could leverage early detection of emotional states for various benefits.
“Although this technology holds the promise of substantial advantages, it is crucial to consider the potential risks, particularly concerning privacy and data misuse,” Jekauc emphasized. “Our research strictly adhered to existing ethical guidelines and data protection regulations. Looking ahead to practical applications of this technology, it will be vital to address ethical and legal concerns proactively.”
Read the original article on: Science Daily
Read more: Kilnam Chon Predicts AI Surpassing Human Intelligence in 30 Years
