The SWELL Knowledge Work Dataset for Stress and User Modeling Research

Koldijk, Saskia; Sappelli, Maya; Verberne, Suzan; Neerincx, Mark A.; Kraaij, Wessel

doi:10.1145/2663204.2663257

Cited by 164 publications

(96 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We used the mean amplitude, skewness, and kurtosis [26] and the six statistical values of the signal for the blood volume pulse (BVP). Finally, we calculated the RMSSD [27], very low (0.05-0.15 Hz) and low (0.15-0.4 Hz) frequency bands and their ratio (low/very low), and the six statistical values for HR.…”

Section: Discussionmentioning

confidence: 99%

Stress Recognition in Daily Work

Nakashima

Kim

Flutura

et al. 2016

Communications in Computer and Information Science

View full text Add to dashboard Cite

Abstract. Automatic detection of work-related stress has attracted an increasing amount of attention from researchers from various disciplines and industries. An experiment is discussed in this paper that was designed to evaluate the efficacy of multimodal sensor measures that have often been used but not yet been systematically tested and compared with each other in previous work, such as pressure distribution sensor, physiological sensors, and an eye tracker. We used the Stroop test and information pick up task as the stressors. In the subject independent case in particular, signals from the combined (chair and floor) pressure distribution sensors, which we consider the most feasible sensors in the office environment, resulted in higher recognition accuracy rates than the physiological or eye tracker signals for the two stressors.

show abstract

Section: Discussionmentioning

confidence: 99%

Stress Recognition in Daily Work

Nakashima

Kim

Flutura

et al. 2016

Communications in Computer and Information Science

View full text Add to dashboard Cite

show abstract

“…We use two public datasets, SWELL [10] and AMIGOS [11]. The SWELL dataset was collected from 25 participants with the aim of understanding the mental stress and emotional attributes of employees in a typical office environment under different working conditions.…”

Section: Datasetsmentioning

confidence: 99%

“…In this paper we propose ECG-based emotion recognition using multi-task self-supervised learning for the first time. We use two publicly available datasets, SWELL [10] and AMI-GOS [11]. First, to train our network with automatically generated labels, we perform 6 different signal transformation tasks.…”

Section: Introductionmentioning

confidence: 99%

Self-Supervised Learning for ECG-Based Emotion Recognition

Sarkar

Etemad

2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

108

View full text Add to dashboard Cite

We present an electrocardiogram (ECG) -based emotion recognition system using self-supervised learning. Our proposed architecture consists of two main networks, a signal transformation recognition network and an emotion recognition network. First, unlabelled data are used to successfully train the former network to detect specific pre-determined signal transformations in the self-supervised learning step. Next, the weights of the convolutional layers of this network are transferred to the emotion recognition network, and two dense layers are trained in order to classify arousal and valence scores. We show that our self-supervised approach helps the model learn the ECG feature manifold required for emotion recognition, performing equal or better than the fully-supervised version of the model. Our proposed method outperforms the state-of-the-art in ECG-based emotion recognition with two publicly available datasets, SWELL and AMIGOS. Further analysis highlights the advantage of our self-supervised approach in requiring significantly less data to achieve acceptable results.

show abstract

“…We used two stress datasets to conduct this study. The first dataset -the SWELL dataset [58] -was collected at the Radboud University. This dataset is a result of experiments conducted on 25 subjects doing office work (for example writing reports, making presentations, reading e-mail and searching for information) who were exposed to quintessential work stressors (e.g., being unexpectedly interrupted by an urgent e-mail and pressure to complete work in a limited time).…”

Section: E T H O D S 21 Stress Datasetsmentioning

confidence: 99%

Effect of Person-specific Biometrics in Improving Generic Stress Predictive Models

Nkurikiyeyezu¹,

Yokokubo²,

Lopez³

2020

Sensors and Materials

View full text Add to dashboard Cite

Because stress is subjective and is expressed differently from one person to another, generic stress prediction models (i.e., models that predict the stress of any person) perform crudely. Only person-specific ones (i.e., models that predict the stress of a preordained person) yield reliable predictions, but they are not adaptable and costly to deploy in real-world environments. For illustration, in an office environment, a stress monitoring system that uses person-specific models would require collecting new data and training a new model for every employee. Moreover, once deployed, the models would deteriorate and need expensive periodic upgrades because stress is dynamic and depends on unforeseeable factors. We propose a simple, yet practical and costeffective calibration technique that derives an accurate and personalized stress prediction model from physiological samples collected from a large population. We validate our approach on two stress datasets. The results show that our technique performs much better than a generic model. For instance, a generic model achieved only a 42.5% ± 19.9% accuracy. However, with only 100 calibration samples, we raised its accuracy to 95.2%±0.5%. We also propose a blueprint for a stress monitoring system based on our strategy, and we debate its merits and limitation. Finally, we made public our source code and the relevant datasets to allow other researchers to replicate our findings.

show abstract

The SWELL Knowledge Work Dataset for Stress and User Modeling Research

Cited by 164 publications

References 15 publications

Stress Recognition in Daily Work

Stress Recognition in Daily Work

Self-Supervised Learning for ECG-Based Emotion Recognition

Effect of Person-specific Biometrics in Improving Generic Stress Predictive Models

Contact Info

Product

Resources

About