Human Experience in Regulated Offices (HERO) dataset

Synopsis: The multimodal Human Experience in Regulated Offices (HERO) dataset captures multimodal human responses to controlled indoor thermal environments during realistic office-like work. Twenty-four healthy adults (pseudonymized as P01–P24) each attended up to two laboratory sessions under distinct thermal regimes: a lower-temperature (LT) condition (~22 °C ± 0.5 °C) and a higher-temperature (HT) condition (~30 °C ± 0.5 °C). In each session (ID_Session = PXX_LT or PXX_HT), participants completed a structured nine-phase protocol comprising baseline, reading, writing, group discussion, simulated phone call, and interleaved washouts. A global timeline file (timeline.csv) specifies the ISO-8601 start and end times of all phases for every session and serves as the temporal backbone for the dataset. This Zenodo release focuses on Level-0 data: device-level streams and context, restricted to the experimental phases and annotated with phase metadata, but without further preprocessing such as filtering, artefact removal, or re-referencing. The dataset includes the following modalities: Electroencephalography, EEG (Muse S): Four-channel scalp EEG (TP9, AF7, AF8, TP10), sampled at ≈256 Hz. Electrodermal Activity, EDA (EmotiBit): Skin conductance recorded from a wearable EmotiBit sensor at ≈15 Hz. Peripheral Skin Temperature, ST (EmotiBit): Peripheral skin temperature (°C) from the same EmotiBit device at ≈7.5 Hz. Environmental indoor climate (env): Experiment room-level environmental measurements recorded throughout the sessions, typically at ≈2 Hz. Perceptual / survey data (survey): Phase-aligned self-report measures acquired at baseline and during/after key phases. The MuSIC_IRP1_HERO dataset is designed to support research on: Neurophysiological dynamics thermal comfort and thermal stress, affective and cognitive state under thermal load, multimodal signal processing and data fusion, and occupant-centric building and indoor climate design. Researchers can use the Level-0 streams to implement their own preprocessing pipelines (e.g. filtering and ICA for EEG, tonic/phasic decomposition for EDA, environmental feature engineering, survey-physiology fusion) while benefiting from consistent temporal alignment across all modalities and experimental phases. All participant identifiers are pseudonymized, and no directly identifying personal information is included. Data collection was conducted under appropriate ethical approvals, and participants provided informed consent for anonymized data sharing for research purposes. The dataset is released for non-commercial research and educational use under a Creative Commons license (see the Zenodo record for the exact license and any additional data-use conditions).