Problem
My engineering thesis (August 2024) is titled "Desarrollo de ambientes virtuales para inducción de diferentes emociones y su efecto en variables fisiológicas", in English development of virtual environments to induce different emotions and their effect on physiological variables.
Virtual reality is used in learning, medicine and entertainment, and it is a candidate tool for exposure therapy. For that, emotions induced in a headset need to be strong enough to show up in the body. The goal was to build software that induces suspense and calm, and to check whether the change can be measured in a physiological signal.
The context and the hardware came from the FONDECYT Exploración project 13220040, "Emocrea", on discovering and recreating emotions in virtual reality with neuroscience and artificial intelligence techniques.
Method
- Platform. An Oculus Quest 2 headset integrated with Unity, with the XR interaction setup and a locomotion system so the participant can move through the scene.
- Two contrasting environments. Suspense uses black and red, dramatic lighting, unsettling sound effects and an eye-occlusion animation. Calm uses white and pastel tones, abundant steady light, relaxing music and no visual restrictions. Both were assembled from modular Unity asset packs.
- Measurement. A photoplethysmography pulse sensor read through an Arduino gave the heart rate.
- Experiment. Five volunteers aged 20 to 25 spent 3 minutes in each environment.
- Analysis in Python. Line plots, scatter plots, heat maps, comparison tables, the Self-Assessment Manikin (SAM) scale and Russell's arousal-valence model.
Results
Heart rate was higher in the suspense environment for all five volunteers. Self-reported arousal went up and valence went down, which is the pattern the arousal-valence model predicts for tension.
| Volunteer | Calm, BPM | Suspense, BPM |
|---|---|---|
| 1 | 78.5 | 95.9 |
| 2 | 81.5 | 100.9 |
| 3 | 79.4 | 93.5 |
| 4 | 74.9 | 90.0 |
| 5 | 78.0 | 91.7 |
| Average | 78.2 | 94.4 |
The average rise was 16 BPM, about 21 percent.
My contribution
I designed the experiment and built the system end to end: the Oculus integration in Unity, the design of the two environments from existing asset packs, the pulse-sensor acquisition with Arduino and the analysis in Python. The environments are not original 3D art; the work was in choosing and combining their parameters so that each induces the intended emotion.
Limits
Five volunteers is a preliminary sample, and heart rate is a single signal. The calm environment did not relax participants as deeply as expected, possibly because of discomfort or dizziness in the headset. Reading the sensor from inside Unity while running the environment hurt performance too much, so that integration needs more work. The thesis proposes adding skin conductance and EEG, and using AI to adapt the stimulus to each person.