Physiological signals from three wearable devices recorded in real-world conditions

Description

This dataset contains multimodal physiological signals collected using wearable devices in two real-world data acquisition scenarios designed to evaluate device performance, signal quality, and practical deployment constraints.

In the first scenario, data were recorded using three wearable devices: Empatica E4, Shimmer3 GSR+, and EmotiBit. This scenario was designed as a multi-device recording setup, enabling direct comparisons across platforms under the same real-world acquisition conditions.

In the second scenario, recordings focused on placement-dependent effects, using a subset of the devices (Simmer3 GSR+ and EmotiBit) to acquire physiological signals from different body locations. This scenario was designed to assess how sensor placement influences signal characteristics and robustness in realistic usage conditions.

Across both scenarios, the dataset includes the full set of signals available from each device. While our study primarily focuses on EDA, PPG, and PPG-derived cardiac metrics (e.g., heart rate and heart rate variability), the raw exports and parsed channels are provided to facilitate alternative analyses and reproducibility.

The dataset was collected to support a comparative study of wearable devices, sensor modalities, and body placements under realistic usage conditions. All signals were processed using a unified and reproducible signal-processing pipeline (https://github.com/gadea-lucas/wearables-signal-comparison) and are intended for research purposes, including wearable sensing evaluation, psychophysiological signal analysis, and affective computing.

Folder structure

The uploaded archive contains a single top-level folder data/ with two subfolders corresponding to the two acquisition scenarios described above.

Note: Data are organized by scenario and device, and grouped by anonymized participant IDs (e.g., 13, 14, …). Individual recordings are provided as timestamped files inside each participant folder.

Scenario 1 (multi-device comparison): data/first/

• data/first/empatica/
  Contains one .csv file per participant (e.g., 13.csv, 14.csv, …) with Empatica E4 signals recorded during the experiment.

• data/first/shimmer/
  Contains one .csv file per participant (e.g., 13.csv, 14.csv, …) with Shimmer3 GSR+ signals recorded during the experiment.

  • raw/: raw stream exports (*.csv) and device information files.

  • parsed/: channel-wise files extracted from the raw stream (e.g., *_EA.csv for EDA, *_PG.csv for PPG, *_HR.csv for heart rate, accelerometer and gyroscope axes), as well as time synchronization files (*_timesyncs.csv, *_timeSyncMap.csv).

    data/first/emotibit/<participant_id>/
    Contains EmotiBit signals recordings for each anonymized participant, including raw exports and processed data:

     

• data/first/Stamps/
  Contains one .txt file per participant with time-stamps of experimental events for this scenario. These markers enable temporal alignment of signals across devices and support event-related analyses.

Scenario 2 (placement-focused recordings): data/second/

• data/second/emotibit_dorsal/<participant_id>/
  EmotiBit recordings acquired at the wrist dorsal placement. Each participant folder contains:

  • raw/: raw EmotiBit exports for each recording session.

  • parsed/: channel-wise files (EDA, PPG, heart rate–related streams, motion axes) and time synchronization files.

• data/second/emotibit_volar/<participant_id>/
  EmotiBit recordings acquired at the wrist volar placement, following the same structure as the dorsal recordings.

• data/second/shimmer_wrist/
  Per-participant .csv files containing Shimmer3 GSR+ recordings acquired at the Empatica-equivalent placement (i.e., GSR at wrist volar position while PPG at wrist dorsal position).

• data/second/shimmer_fingers/
  Per-participant .csv files containing Shimmer3 GSR+ recordings acquired at the fingers placement (as recommended in the device manual).

• data/second/Stamps/
  Contains one .txt file per participant with time-stamps of experimental events for Scenario 2, enabling temporal alignment and event-based analysis within this placement-focused setting.