How the Dataset is structured.
The StressID Dataset contains:
- physiological recordings for 65 subjects,
- video recordings for 54 participants, and
- audio for 56 participants.
Following data collection, we split each recorded session into individual tasks: one 3 minutes breathing recording (block 1), 2 recordings corresponding to the watching of the video-clips (block 2) of respectively 2 and 3 minutes, 7 separate 1 minute recordings of the interactive tasks (block 3), and a 5 minutes relaxation recording (block 4).
After splitting, StressID consists of approximately 1119 minutes of annotated physiological signals recordings, 918 minutes of video recordings, and 385 minutes of audio.
Dataset composition.
| Task/Stressor | Count physiological (min) | Count video (min) | Count audio (min) |
| Breathing | 65 (195) | 52 (156) | 0 (0) |
| Video 1 | 64 (185) | 52 (150) | 0 (0) |
| Video 2 | 64 (126) | 53 (104) | 0 (0) |
| Counting 1 | 65 (65) | 54 (54) | 55 (55) |
| Counting 2 | 65 (65) | 54 (54) | 55 (55) |
| Stroop | 65 (65) | 54 (54) | 55 (55) |
| Speaking | 65 (65) | 54 (54) | 55 (55) |
| Math | 65 (65) | 54 (54) | 55 (55) |
| Reading | 65 (65) | 54 (54) | 55 (55) |
| Counting 3 | 65 (65) | 54 (54) | 55 (55) |
| Relax | 63 (158) | 52 (130) | 0 (0) |
| Total | 711 (1119) | 587 (918) | 385 (385) |
More technical information can be found here.
