Home // SIGNAL 2023, The Eighth International Conference on Advances in Signal, Image and Video Processing // View article
Authors:
Vasileios-Rafail Xefteris
Athina Tsanousa
Spyridon Symeonidis
Sotiris Diplaris
Francesco Zaffanela
Martina Monego
Maria Pacelli
Stefanos Vrochidis
Ioannis Kompatsiaris
Keywords: Stress level detection, wearable sensors, smart vest, multimodal fusion
Abstract:
Stress as a mental/physiological reaction of a person in a challenging situation of high discomfort can affect his/her ability to focus and perform fast and accurate decisions. Thus, stress can be a key factor in cases of emergency, when first responders need to be fast and accurate. Continuous monitoring of the stress levels of the first responders can be crucial in cases of disaster management situations. Wearable devices and physiological sensors provide real-time monitoring of physiological signals, which can be helpful for real-time stress monitoring. This work describes the stress detection module of the xR4DRAMA project and the results of its application during a disaster management pilot scenario. For this cause, a wearable smart vest equipped with an electrocardiograph (ECG) sensor, respiration (RSP) sensor, and an Inertial Measurement Unit (IMU) with an accelerometer, gyroscope, magnetometer, and quaternion sensors has been used. An initial data collection was performed to train the stress detection module, and the trained model was deployed for real-time stress detection of first responders in the pilot scenario. The training performed includes a massive feature extraction from the different modalities, and the test of four machine learning algorithms and six fusion and three feature selection techniques. The results of the continuous valued stress levels detection indicate that the best performing combination is the eXtreme Gradient Boosting (XGB) algorithm with the use of a Genetic Algorithm (GA) feature selection technique, achieving a Mean Square Error (MSE) of 0.0567. Results from the pilot show that the stress level detection module can operate in real time in real life conditions, offering reasonable results regarding the detected stress levels.
Pages: 7 to 12
Copyright: Copyright (c) IARIA, 2023
Publication date: March 13, 2023
Published in: conference
ISSN: 2519-8432
ISBN: 978-1-68558-057-5
Location: Barcelona, Spain
Dates: from March 13, 2023 to March 17, 2023