REVELIO

Multimodal physiological and behavioral recordings for task load estimation during an n-back test and two commercial video games (Hogwarts Legacy, Overcooked! 2). This record holds the Public Set: 30 of the 45 REVELIO gaming sessions.

How to cite Please cite this dataset record and both papers: ADABase introduced the acquisition setup, and REVELIO extends it with the gaming cohort.

@articleoppelt2023adabase, author = Oppelt, Maximilian P. and Foltyn, Andreas and Deuschel, Jessica and Lang, Nadine R. and Holzer, Nina and Eskofier, Bjoern M. and Yang, Seung Hee, title = ADABase: A Multimodal Dataset for Cognitive Load Estimation, journal = Sensors, year = 2023, volume = 23, number = 1, pages = 340, doi = 10.3390/s23010340

@articleoppelt2026revelio, author = Oppelt, Maximilian P. and Foltyn, Andreas and Lang-Richter, Nadine R. and Eskofier, Bjoern M., title = REVELIO: Multimodal Task Load Estimation for Cross-Domain Generalization, journal = International Journal of Human–Computer Interaction, publisher = Taylor & Francis, year = 2026, doi = 10.48550/arXiv.2509.01642, note = In press. DOI refers to the arXiv preprint arXiv:2509.01642

Dataset overview Participants. ADABase and REVELIO together comprise 96 sessions from 88 subjects (51 in the driving study, 37 in the gaming study). Five people took part in both studies, so the subjects are 83 different individuals. The sessions are split into four Zenodo records:

Subset Sessions DOI

ADABase: Cognitive Driving Public Set 30 10.5281/zenodo.24051552

ADABase: Cognitive Driving Private Set 21 10.5281/zenodo.24059812

REVELIO: Cognitive Gaming Public Set (this record) 30 10.5281/zenodo.24060698

REVELIO: Cognitive Gaming Private Set 15 10.5281/zenodo.24076078

The REVELIO gaming cohort comprises 45 sessions from 37 participants.

Tasks. Each session contains an n-back test and two commercial video games, Hogwarts Legacy and Overcooked! 2, each with baseline/training and task segments. Participants rated their task load with the NASA-TLX after each task segment. A session lasted about 2–3 hours. See the REVELIO paper for the protocol.

Modalities. The sensors, recording equipment and n-back protocol are the same as in ADABase.

Physiological signals (BIOPAC MP160, recorded at 2000 Hz, stored at 1 kHz): ECG, PPG, EDA, EMG (trapezius), respiration, skin temperature Eye tracking (Tobii Pro Fusion, 250 Hz): gaze and pupil diameter Behavioral features extracted from facial video (camera at 25 Hz): facial action units, head pose, body landmarks. No raw video or images are released. Performance measures, NASA-TLX ratings, and participant meta-information (e.g. personality traits, gaming experience)

Public/private split. The 45 sessions are split into two Zenodo records:

Public Set: 30 sessions, sampled at random, access on request. Private Set: 15 sessions, retained as a hidden test set.

Data collection was approved by the ethics committee of Friedrich-Alexander-Universität Erlangen-Nürnberg (no. 129_21 B, 21 April 2021). All participants gave written informed consent, and only participants who consented to data sharing are included.

Files This record contains 30 HDF5 files (25.02 GB in total, see file list), plus checksums.md5.

File naming scheme: -.h5

NNNN: four-digit participant ID. IDs are unique across ADABase and REVELIO, so the same NNNN always denotes the same person. IDs 0051–0082 are new in REVELIO. IDs 0009, 0013, 0029, 0031 and 0045 belong to participants who also took part in ADABase. CODE: pseudonym of the session. Participants with two sessions have two files with the same NNNN and different CODEs.

Example: 0013-SKR11.h5 and 0013-YAF19.h5 are the two gaming sessions of participant 0013.

All 30 files 0009-WMS15.h5, 0013-SKR11.h5, 0013-YAF19.h5, 0029-GPW07.h5, 0031-GBR27.h5, 0045-GGZ24.h5, 0045-LJT10.h5, 0051-CEN35.h5, 0052-COA12.h5, 0053-DBE20.h5, 0054-DFI08.h5, 0055-EAB20.h5, 0056-EEE10.h5, 0056-YWK22.h5, 0057-FHC08.h5, 0057-SGF31.h5, 0058-FIL28.h5, 0059-IUS14.h5, 0059-KNC19.h5, 0060-KBG08.h5, 0061-KOB08.h5, 0062-LMI20.h5, 0063-QGP09.h5, 0064-QIK35.h5, 0065-SKW18.h5, 0065-ZJC18.h5, 0066-VYM28.h5, 0067-ZAT15.h5, 0068-ZAU14.h5, 0069-ZYP27.h5

HDF5 structure Each file is written with pandas (PyTables) and holds four tables, which pandas.read_hdf reads directly. Driving (ADABase) and gaming (REVELIO) files share this layout.

Key Rows Content

SIGNALS one per millisecond All time series on a common 1 kHz grid. The index is the time in ms since the start of the recording. Annotation columns STUDY, PHASE, LEVEL, then one column per channel (see below).

SUBJECTIVE one per rated level NASA-TLX ratings with STUDY, PHASE, LEVEL. Six dimensions MENTAL, PHYSICAL, TEMPORAL, PERFORMANCE, EFFORT, FRUSTRATION on a 21-point scale (0 to 20), and their pairwise-comparison weights WEIGHT (sum 15).

PERFORMANCE one per scored level Task performance with STUDY, PHASE, LEVEL. The columns differ between tasks and are empty where they do not apply, e.g. F1, VISUAL F1, AUDITIVE F1, REACTION TIME, VISUAL MEAN REACTION TIME, AUDITIVE MEAN REACTION TIME, SONGS RECALL, RACE TIME (text, mm:ss,cc), MISSED RINGS, SCORE. Some levels have no score.

META one per session Demographics and questionnaires, with a two-level column index grouped by topic, e.g. ("Subject", "Age"), ("Subject", "Sex").

Channels in SIGNALS.

Columns Content

RAW_ECG_I, RAW_ECG_II ECG, two leads

RAW_PPG Photoplethysmogram

RAW_EDA Electrodermal activity

RAW_RSP Respiration (chest belt)

RAW_EMG EMG (trapezius)

RAW_SKT Skin temperature

ECG_I_PEAKS, PPG_PEAKS, RSP_PEAKS Event markers: R peaks in ECG I, pulse peaks in PPG, inspiration peaks in respiration

EYE_LEFT_*, EYE_RIGHT_* Eye tracker, e.g. EYE_LEFT_PUPIL_DIAMETER (mm), EYE_LEFT_PUPIL_VALIDITY, EYE_LEFT_GAZE_POINT_ON_DISPLAY_AREA_X

FACE_AU* Facial action units extracted from video, e.g. FACE_AU12

MOVE_* Body landmarks extracted from video, e.g. MOVE_X_NOSE

The full channel list is printed by the example in "Loading the data in Python".

Sampling rates. The biosignals are stored at 1 kHz. Streams with a lower native rate (eye tracker 250 Hz, video features 25 Hz) are NaN between their samples. Calling .dropna() on a column recovers the native samples. The row number equals the time in ms, so the start and stop arguments of pandas.read_hdf select a time range without reading the whole file. A file has about 9 million rows, so read only the columns and time range you need.

Annotation. Every sample in SIGNALS carries three labels:

STUDY: the task. Time outside all tasks is n/a. PHASE: baseline, train (practice), test, or n/a between phases. LEVEL: numbers the conditions within a study and phase. In the gaming files, STUDY and PHASE are stored as pandas categories, which pandas.read_hdf decodes automatically.

Verbatim fromZenodo· Licence: see record