DeStress: Deep Learning for Unsupervised Identification of Mental Stress in Firefighters from Heart-Rate Variability (HRV) Data

Oskooei, Ali; Chau, Sophie Mai; Weiß, Jonas; Sridhar, Arvind; Martínez, María Rodríguez; Michel, Bruno

doi:10.1007/978-3-030-53352-6_9

Cited by 31 publications

(26 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this scenario, an unobtrusive approach like the one presented in this article may provide insights about the decisions made by consumers and their reactions to certain products. Alternatively, for professionals working under stressful conditions such as firefighters, industrial workers, or surgeons, stress quantitative assessment may be useful during training (Chrouser et al, 2018;Oskooei et al, 2021). Thereby, individual stress models of the trainees could provide the instructors valuable stress biofeedback that may be used to support education techniques.…”

Section: Discussionmentioning

confidence: 99%

Quantitative Assessment of Stress Through EEG During a Virtual Reality Stress-Relax Session

Pérez-Valero

Vaquero-Blasco

López-Gordo

et al. 2021

Front. Comput. Neurosci.

View full text Add to dashboard Cite

Recent studies have addressed stress level classification via electroencephalography (EEG) and machine learning. These works typically use EEG-based features, like power spectral density (PSD), to develop stress classifiers. Nonetheless, these classifiers are usually limited to the discrimination of two (stress and no stress) or three (low, medium, and high) stress levels. In this study we propose an alternative for quantitative stress assessment based on EEG and regression algorithms. To this aim, we conducted a group of 23 participants (mean age 22.65 ± 5.48) over a stress-relax experience while monitoring their EEG. First, we stressed the participants via the Montreal imaging stress task (MIST), and then we led them through a 360-degree virtual reality (VR) relaxation experience. Throughout the session, the participants reported their self-perceived stress level (SPSL) via surveys. Subsequently, we extracted spectral features from the EEG of the participants and we developed individual models based on regression algorithms to predict their SPSL. We evaluated stress regression performance in terms of the mean squared percentage error (MSPE) and the correlation coefficient (R2). The results yielded from this evaluation (MSPE = 10.62 ± 2.12, R2 = 0.92 ± 0.02) suggest that our approach predicted the stress level of the participants with remarkable performance. These results may have a positive impact in diverse areas that could benefit from stress level quantitative prediction. These areas include research fields like neuromarketing, and training of professionals such as surgeons, industrial workers, or firefighters, that often face stressful situations.

show abstract

Section: Discussionmentioning

confidence: 99%

Quantitative Assessment of Stress Through EEG During a Virtual Reality Stress-Relax Session

Pérez-Valero

Vaquero-Blasco

López-Gordo

et al. 2021

Front. Comput. Neurosci.

View full text Add to dashboard Cite

show abstract

“…KNN was found to be the most effective model with over 90% accuracy [11]. Oskooei et al [12] suggested an unsupervised method for binary stress classification using heart-rate variability (HRV) extracted through CAE. Sarkar and Etemad [23] proposed a self-supervised ECG learning for classifying emotion arousal, valence and stress by first training a multi-tasking CNN using transformed signals and then transferring the network for emotion prediction task.…”

Section: Related Prior Workmentioning

confidence: 99%

“…Based on [42] and random search, the matching network was trained using 70% of the samples with a batch size of 128 for 400 epochs. Cosine similarity (L cs ), as shown in Equation (12), was selected as the matching function to measure the distance between the prediction and the ground truth.…”

Section: A Experimental Setup 1) Stress Datasetmentioning

confidence: 99%

“…Recent studies in the area of stress monitoring have led to machine learning (ML) models for stress evaluation and classification. Aqajari et al [11] and Oskooei et al [12] proposed effective ways of physiological signal feature extraction through neural networks such as Convolutional Neural Networks (CNN) and Convolutional Autoencoders (CAE). Others such as Tripathi and Beigi [13], Busso et al [14] and Jaiswal et al [15] proposed to use images and audio information captured from recorded videos with predesigned scenes for stress classification.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Resilient and Hierarchical IoT-based Solution for Stress Monitoring in Everyday Settings

Jiang¹,

Firouzi²,

Chakrabarty³

et al. 2021

Preprint

View full text Add to dashboard Cite

<div><div><div><p>Long-term stress is a global health concern because it impacts our physical and mental health. The emergence of Internet of Things (IoT) and Artificial Intelligence (AI) makes stress monitoring and treatment more accessible compared to today’s physician-centered healthcare system. However, existing solutions either fail to incorporate IoT technology or are not cost-effective. We propose a resilient, hierarchical IoT-based solution for stress monitoring to tackle the above problems. Multimodal data was collected from wearable sensors and underwent preprocessing, feature extraction, and multiple imputation. We applied three feature-selection methods prior to lightweight SVM classification at the edge layer, and utilized a CNN and a matching network model in the cloud layer. We obtained an accuracy of 86.7347% and an F1 score of 0.8725 at the edge using only 10 features selected based on the Fisher score. An accuracy of 98.9247% and an F1 score of 0.9876 was achieved by a matching network model based on electrocardiogram (ECG) data. The trade-off between the communication cost from the edge to the cloud and the overall accuracy was evaluated. Our hierarchical-IoT solution for stress-level evaluation provides insights into the potentiality of IoT and AI technology-based eHealth solutions.</p></div></div></div>

show abstract

“…Machine learning (ML) has shown very promising results in many studies investigating data from firefighters or fNIRS tools. For instance, several studies have used ML algorithms, such as decision trees (DT), k-nearest-neighbors (KNN), and support vector machines (SVM), with physiological data such as Heart Rate Variability (HRV), body temperature, and behavior tracking sensors such as accelerometers, to detect mental workload, exertion, and stress in firefighters [ 29 , 30 , 31 ]. Similarly, other studies have used ML algorithms with other populations to detect task difficulty, mental workload, fatigue, engagement, enjoyment, and user performance [ 32 , 33 , 34 , 35 , 36 , 37 , 38 ].…”

Section: Introductionmentioning

confidence: 99%

Brain Activity-Based Metrics for Assessing Learning States in VR under Stress among Firefighters: An Explorative Machine Learning Approach in Neuroergonomics

et al. 2021

View full text Add to dashboard Cite

The nature of firefighters` duties requires them to work for long periods under unfavorable conditions. To perform their jobs effectively, they are required to endure long hours of extensive, stressful training. Creating such training environments is very expensive and it is difficult to guarantee trainees’ safety. In this study, firefighters are trained in a virtual environment that includes virtual perturbations such as fires, alarms, and smoke. The objective of this paper is to use machine learning methods to discern encoding and retrieval states in firefighters during a visuospatial episodic memory task and explore which regions of the brain provide suitable signals to solve this classification problem. Our results show that the Random Forest algorithm could be used to distinguish between information encoding and retrieval using features extracted from fNIRS data. Our algorithm achieved an F-1 score of 0.844 and an accuracy of 79.10% if the training and testing data are obtained at similar environmental conditions. However, the algorithm’s performance dropped to an F-1 score of 0.723 and accuracy of 60.61% when evaluated on data collected under different environmental conditions than the training data. We also found that if the training and evaluation data were recorded under the same environmental conditions, the RPM, LDLPFC, RDLPFC were the most relevant brain regions under non-stressful, stressful, and a mix of stressful and non-stressful conditions, respectively.

show abstract

DeStress: Deep Learning for Unsupervised Identification of Mental Stress in Firefighters from Heart-Rate Variability (HRV) Data

Cited by 31 publications

References 34 publications

Quantitative Assessment of Stress Through EEG During a Virtual Reality Stress-Relax Session

Quantitative Assessment of Stress Through EEG During a Virtual Reality Stress-Relax Session

A Resilient and Hierarchical IoT-based Solution for Stress Monitoring in Everyday Settings

Brain Activity-Based Metrics for Assessing Learning States in VR under Stress among Firefighters: An Explorative Machine Learning Approach in Neuroergonomics

Contact Info

Product

Resources

About