PhySF, Physiological Sense of Flow

Description

PhySF (“Physiological Sense of Flow”) is a multimodal wearable-sensor recording assembled to study the psychological flow state — the absorbed, effortless engagement that arises when a person is fully immersed in a challenging-but-tractable activity. Each subject is recorded simultaneously across three consumer wearables (a 14-channel Emotiv EPOC X EEG headset, an Empatica E4 wrist sensor, and a Biosignalplux RespiBAN respiration belt) while performing arithmetic and reading tasks designed to induce either a flow or a non-flow (no-flow) condition (Irshad et al., 2023). The corpus was collected by the group of Marcin Grzegorzek at the Institute of Medical Informatics, University of Lübeck, in collaboration with the Department of Psychology at Lübeck and partner psychology/economics groups at the University of Silesia in Katowice, the University of Economics in Katowice, and Jan Kochanowski University of Kielce. It was used to train machine-learning classifiers that discriminate flow from non-flow from physiological signals, and to test whether transfer learning from emotion-recognition tasks (DEAP) improves flow detection. The dataset comprises 25 subjects recorded in the two conditions, yielding 46 EEG CSV files; participant demographics (age range, sex breakdown) are not documented in the publicly available material.

Citation

• Paper DOI: 10.1016/j.compbiomed.2023.107489
• Dataset DOI: TODO(confirm)
• Reference: Irshad, M. T., Li, F., Nisar, M. A., et al. (2023). Wearable-based human flow experience recognition enhanced by transfer learning methods using emotion data. Computers in Biology and Medicine 166, 107489.

Catalog citation block

Irshad, M. T., Li, F., Nisar, M. A., Huang, X., Buss, M., Kloep, L., Peifer, C., Kozusznik, B., Pollak, A., Pyszka, A., Flak, O., & Grzegorzek, M. (2023). Wearable-based human flow experience recognition enhanced by transfer learning methods using emotion data. Computers in Biology and Medicine, 166, 107489. doi:10.1016/j.compbiomed.2023.107489

Download and access

Field	Value
Official source	TODO(confirm)
Access class	request
Size	TODO(confirm)
Format	csv
BIDS	false
Mirrors	none recorded

Access steps:

1. Contact the authors of the Computers in Biology and Medicine article (DOI 10.1016/j.compbiomed.2023.107489)

Source & access

The dataset underlies the data-analysis paper Irshad et al. (2026), “Wearable-based human flow experience recognition enhanced by transfer learning methods using emotion data,” Computers in Biology and Medicine, vol. 166, article 107489 (doi:10.1016/j.compbiomed.2023.107489). No dedicated public repository (OpenNeuro / Zenodo / OSF / Kaggle / PhysioNet) accession or DOI was located for the raw recordings, and the dataset does not appear in recent systematic reviews of open physiological-signal datasets — it is most likely available on request from the authors. License is not documented. The research was funded by the Deutsche Forschungsgemeinschaft (DFG grant 465142069, project “Team-Flow und Team-Effektivität in virtuellen Teams” led by Grzegorzek and Peifer), the University of Lübeck (22-112), and the Polish Narodowe Centrum Nauki (2020/39/G/HS6/02124).

Data structure

The dataset is delivered as flat CSV files (one per subject-condition recording), not in BIDS. Each CSV holds 23 columns: 14 EEG (Emotiv EPOC X montage) + 5 Empatica E4 + 4 RespiBAN, i.e. all three device streams are merged column-wise into a single per-recording table. No derivatives, git-annex, or DataLad structure is present — this is a raw CSV drop.

The 14 EEG channels (Emotiv EPOC X 10-20 montage) are: AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4.

License

• Name: none
• Note (verbatim from the directory): license undocumented; no public repository located — data on request from the authors
• Snippets on this site: no — the site ships no snippet or precomputed product from this dataset; lessons link the official access page.

Acquisition

Field	Value
Device	14-channel wireless saline-electrode headset (Emotiv EPOC X) + Empatica E4 wrist sensor + Biosignalplux RespiBAN respiration belt
Device class	consumer
Channels	14
Sampling rate (Hz)	128
Online filters	device hardware default (≈ 0.16–43 Hz; 50/60 Hz notches) per the manufacturer specification
Reference	CMS/DRL at P3/P4 per the manufacturer specification
Mains frequency (Hz)	TODO(confirm)
Paradigms	flow-task, mental-arithmetic, reading
Participants	25
Population	25 adults; demographics undocumented
Clinical groups	none
Sessions	TODO(confirm)
Durations	flow and non-flow recordings per subject; per-recording duration not documented

Acquisition

EEG was acquired with an Emotiv EPOC X headset — 14 channels, saline felt sensors, with CMS/DRL references at P3/P4 (left/right mastoid as an alternative reference) per the manufacturer specification (EMOTIV EPOC X product documentation). The device samples internally at 2048 Hz per channel (single sequential ADC) and downsamples to the stored 128 samples/s (a 256 SPS mode is also user-selectable; the PhySF files use 128 Hz). Resolution is 0.51 µV/LSB in 14-bit mode (0.1275 µV in 16-bit mode). A built-in 5th-order Sinc filter gives a hardware bandwidth of 0.16—43 Hz, with hardware digital notch filters at both 50 Hz and 60 Hz for mains rejection. Recorded alongside the EEG, per simultaneous session:

• Empatica E4 wristband — 5 channels (blood-volume pulse / PPG, electrodermal activity, skin temperature, and accelerometry), stored as a 5-column block.
• Biosignalplux RespiBAN (respiration, belt) — 4 channels.

Per-recording duration is not documented in the available sources.

Participants

The dataset contains physiological recordings from 25 subjects (Irshad et al., 2023). Each subject contributes both a flow and a non-flow recording. The published material does not report participant age range/mean, sex distribution, or recruitment/inclusion criteria. The participants are described only as healthy adults performing cognitive (arithmetic and reading) tasks — no clinical group is involved.

Tasks / conditions

Two conditions are recorded per subject — a flow condition and a non-flow (no-flow) condition — giving the 46-file count (flow + no-flow across 25 subjects, with two files unaccounted for, likely a dropped or partial recording). Flow vs non-flow was manipulated through arithmetic and reading tasks with difficulty calibrated to the subject’s skill, the standard challenge-skill-balance manipulation for inducing flow (Irshad et al., 2023). Condition labels (flow / non-flow) serve as the classification targets in the source study.

Used on this site

L7.7

Loader

Fetches only the stated subset. Recorded loader: TODO(confirm).

TODO(confirm) — the catalog records no loader for this dataset.

Caveats worth knowing

• No public repository, DOI or license text; recording durations and demographics undocumented.

Notable caveats

• Hardware-limited Nyquist. The effective usable band tops out at 43 Hz (hardware-limited by the EPOC X 0.16—43 Hz bandwidth), not the 64 Hz implied by the 128 Hz sampling rate — the EPOC X rolls off above 43 Hz. Gamma analyses are not supported.
• Consumer EEG. EPOC X is a 14-channel saline-sensor consumer headset with CMS/DRL referencing at P3/P4; spatial coverage is sparse (frontal-heavy, no central midline) and signal quality / motion susceptibility are well below clinical gel systems.
• Multi-device alignment. EEG, wristband, and respiration are captured across three independent devices and merged into one CSV; cross-stream temporal alignment relies on the per-file column layout (14 EEG + 5 Empatica + 4 RespiBAN) and the devices’ own clocks rather than a hardware sync trigger.
• No public accession / license. No repository DOI or license could be located; treat provenance as author-supplied. Participant demographics and recording durations are undocumented.

Related datasets

• MultiPENG (EEG stream)
• ArEEG_Words
• EEG for natural-image recognition (VEP)
• EEGEmotions-27
• ERP CORE (Compendium of Open Resources and Experiments)

# Loader: TODO(confirm) — the catalog records no loader for this dataset.
# Download a small subset from the official source: TODO(confirm)