SPACE - Stress and Physiological Assessment in a Controlled Environment

SPACE - Stress and Physiological Assessment in a Controlled Environment Data description Participants A total of 128 participants, consisting of students and staff from Université Laval, took part in this data collection. Data collection took place from February 17 to June 23, 2025. A Fossil Gen 5 smartwatch (model D10F1), a BioHarness Zephyr chest strap, and a millimeter wave (mmWave) radar (Texas Instruments IWR1443) were used to record physiological data. Additionally, videos of participant’s faces were recorded with a Logitech webcam. Audiovisual and location data recorded as part of the experiment are not included in the public dataset due to participant confidentiality. 42 participants were exposed to the Multiple stressors condition, while 42 participants experienced the Social stressor condition. 44 other participants were exposed to the Temporal stressor condition. Participants who start with P0 (e.g., P025) are in the Multiple stressor group. Those who start with P1 (e.g., P125) are in the Social stressor condition group. Participants ID starting with P2 (e.g., P201) are in the Temporal stressor condition. The following participants, listed in the “participants.csv” file, had data collection issues: P026, P029, P117, P119, P121, P202, P203, P221, and P225. The table below describes the data collection issues. Participant ID Data collection issue P026 Invalid data from the Bioharness. P029 Physiological data did not record for the first two sessions. P117 No facial data from the webcam. P119 Physiological data did not record for the first two sessions. P121 Abnormal ECG data. P202 Battery issues with the Fossil smartwatch. P203 The signal nexus failed to record the data. P221 Poor facial data from the webcam. P225 Invalid data from the Bioharness. Data Collection Procedure The data collection process is summarized in the diagram below: Participants began the experiment by completing the Perceived Stress Scale (PSS; Cohen et al., 1983) questionnaire and sociodemographic data such as age, sex, physical condition, etc. Baseline (5 minutes): Participants were asked to remain at rest while physiological signals were recorded. They were instructed to simply gaze at a black fixation cross on a white screen and refrain from any other activity. Following this, participants completed the Visual Analogue Scale for Stress (VAS Stress), a slider ranging from 0 to 10 to assess their perceived stress levels, as well as the NASA-TLX questionnaire (Hart & Staveland, 1988), which evaluates perceived workload. Task Sessions (6 minutes total): Each task consisted of 5 minutes of OpenMATB (MultiAttribute Task Battery; Cegarra et al., 2020), with an interruption at the halfway point (2 minutes and 30 seconds). During this break, participants were given 60 seconds to speak aloud for voice analysis purposes. The task conditions were as follows: Tutorial (6 minutes): Designed to familiarize participants with the tasks and the interface. Easy: The first scenario, designed to be of low difficulty. Stressful: A scenario with more frequent events in the OpenMATB configuration, aiming to create a higher time pressure. In the multiple stressor group, an additional auditory stressor was introduced to further challenge participants. In the social stress group, an evaluator entered at this point, aiming to induce social evaluative threat (Allen et al., 2017). In the Temporal stressor condition, no sound or confederate was present; the stress only stemmed from the task difficulty. Easy_final: The final "easy" scenario, designed to be of low difficulty and conclude the experiment. Data organization Each of the files in the dataset are described below. participants.csv This spreadsheet contains notes on each recording session taken by experimenters. While the notes were originally written in French, two additional columns were added to indicate in English which participants had data collection issues that excluded them from analyses using the physiological and video data. OpenMATB.zip This contains the files necessary to run the OpenMATB task. It includes configuration files for each condition, a video played during the baseline, and an auditory stimulus played during the “multiple stressor” condition. Multiple_stress_nonvideo.zip, Social_stress_nonvideo.zip and Temporal_stress_nonvideo.zip Each of these zip files contain anonymized physiological, behavioral, and questionnaire data for the two groups (multiple stress, social stress and temporal stress). Once extracted, there should be subfolders for each participant called “PXX”. Within each of the participant subfolders, the following data can be found: PXXX.zip This zip file, at the root of the participant subfolder and named according to the participant, contains data downloaded from a dashboard used by the experimenter to manage sessions recorded with the SensorHub mobile application. SensorHub (Gagnon et al., 2014) is a platform that enables multi-sensor integration: electroencephalography, electrocardiography (ECG), accelerometers, location, cameras, temperature, etc. In addition to data received directly from sensors, SensorHub calculates higher-level features (e.g., heart rate variability) and runs machine learning models (e.g., stress) in near-real time. The dashboard data extraction contains the metadata of the session along with the sensor data sent in real-time to the dashboard being used to manage the session. To access the physiological signals, navigate to the "features" subfolder. Session metadata such as the start and stop timestamps can be found in “sessions.json”. While the dashboard data extraction contains the necessary features for the analyses conducted within this project, the local SensorHub data is also included since dashboard data extraction does not include every original raw data stream. Note that dashboard recordings often start a few seconds (or occasionally minutes) after the local SensorHub logs, since phone-based recording is started first to confirm system readiness before the session was started in the dashboard. DATETIME__SH/ This folder contains the SensorHub data logged directly on the phone, as soon as the system was started. Machine learning model outputs and location data were excluded from the public dataset, but the following raw data and features are available: AndroidWatchHeartRaw.csv: Raw heart-rate data recorded by the Android smartwatch, with timestamps and beats-per-minute (BPM) values. BatteryLevelFeature.csv: BioHarness battery level over time, expressed as a percentage at each timestamp. BioAccelerometerRaw.csv: Raw three-axis accelerometer readings from the BioHarness. BioBreathingRaw.csv: Respiration rate data recorded from the BioHarness. BioEcgRaw.csv: ECG data recorded from the BioHarness. ECGLeadFeature.csv: ECG data formatted by SensorHub in a sensor-agnostic format. HeartRateBaselineFeature.csv: Statistical summaries extracted from baseline heart rate data recorded by the BioHarness during a rest period at the beginning of the experiment, specified by the experimenter in the dashboard. HeartRateFeature.csv: Heart rate time series from all sensors in SensorHub format, including per-sample beats per minute and associated metadata such as frame and heart rate confidence (measure of signal quality calculated by each sensor). HeartRateFusionFeature.csv: Fused heart rate estimates combining mmWave, Android Watch, and BioHarness inputs. HeartRateVariability.csv: Heart rate variability (HRV) features computed by SensorHub from R-R intervals recorded by the BioHarness (e.g. low-frequency and high-frequency power) extracted over short (120-second) and long (300-second) time windows. Label.csv: Session labels marking the start of each recording, as logged from SensorHub. MmWaveDriverRaw.csv: Output from the TI IWR1443 mmWave radar (respiration rate and heart rate estimates from raw radar metrics). RespirationBaselineFeature.csv: Statistical summaries of baseline respiration rate data recorded by the BioHarness. RespirationFeature.csv: Respiration rate time series from all sensors in SensorHub format and associated metadata. RespirationFusionFeature.csv: Fused respiratory rate estimates combining mmWave and BioHarness inputs. RRIntervalFeature.csv: Interval between two R-peaks of ECG data, calculated by the BioHarness. SessionEventRaw.csv: Timestamps for when baseline and recording periods began and ended. MATB/ This folder contains data from the five MATB sessions: Session 1: Corresponds to the first questionnaire, following the baseline. To determine the start time of baseline, subtract 5 minutes from the start of the session. Session 2: Represents the tutorial. Session 3: Represents the "easy" scenario. Session 4: Represents the "stressful" scenario. Session 5: Represents the final "easy" scenario. The format for the CSV filenames is as follows: 51_250218_101338 51: Session #51 from MATB. 250218: Date in the format (DDMMYY), i.e., February 18, 2025. 101338: Time of data collection (10:13:38). There are two "easy" scenarios, counterbalanced between participants (easy & easy_final). Reponses_pss.csv This csv file contains data from the Perceived Stress Scale (PSS) questionnaire. PSS is used to assess the level of stress experienced by an individual in everyday life. Reponses_sociodemo.csv This csv file includes sociodemographic data, such as sex, age, physical fitness level, etc. Combined_participants_timestamps.csv The CSV file contains the times and scenarios corresponding to the participants. This will facilitate the mapping of stress episodes to scenarios. Multiple_stress_facial_landmarks.zip, Social_stress_facial_landmarks.zip and Temporal_stress_facial_landmarks.zip Each of these zip files contain anonymized facial landmark images for the two groups (multiple stress, social stress and temporal stress). Once extracted, there should be subfolders for each participant called “PXX”. Within each of the participant subfolders, the landmarks are organized in folders named according to the start time of the first frame in the folder in Unix epoch milliseconds. The landmark images were generated using the dlib library from the webcam video recorded throughout the entire experiment. Note that some frames are missing when landmarks could not be generated, e.g., when the participant’s face was occluded. References Allen, A. P., Kennedy, P. J., Dockray, S., Cryan, J. F., Dinan, T. G., & Clarke, G. (2017). The Trier Social Stress Test: Principles and practice. Neurobiology of Stress, 6, 113-126. https://doi.org/10.1016/j.ynstr.2016.11.001 Cegarra, J., Valéry, B., Avril, E., Calmettes, C., & Navarro, J. (2020). OpenMATB: A MultiAttribute Task Battery promoting task customization, software extensibility and experiment replicability. Behavior Research Methods, 52, 1980-1990. https://doi.org/10.3758/s13428-020-01364-w Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress. Journal of Health and Social Behavior, 24, 385-396. https://doi.org/10.2307/2136404 Gagnon, J.-F., Lafond, D., Rivest, M., Couderc, F., & Tremblay, S. (2014). Sensor-Hub: A realtime data integration and processing nexus for adaptive C2 systems. Proceedings of the Sixth International Conference on Adaptive and Self-Adaptive Systems and Applications, 63-67. Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In P. A. Hancock & N. Meshkati (Eds.), Human mental workload (pp. 139-183). North-Holland. https://doi.org/10.1016/S01664115(08)62386-9 Lesage, F.-X., Berjot, S., & Deschamps, F. (2012). Clinical stress assessment using a visual analogue scale. Occupational Medicine, 62, 600-605, https://doi.org/10.1093/occmed/kqs140 Acknowledgements This work was made possible thanks to the technical support of the Canadian Space Agency (CSA) and financial support of Defence Research and Development Canada (DRDC), as well as research grants awarded to S. Tremblay (National Sciences and Engineering Research Council of Canada, no. RGPIN-2022-04852) and to A. Marois (Fonds de recherche du Québec - Nature et technologies, no. 342553).