Applying Facial Expression Analysis Software to Estimate the User Experience of LEGO building by Analyzing Facial Action Units
Authors
Andersen, Lisa Bondo ; Thygesen, Mathias
Term
4. term
Education
Publication year
2024
Submitted on
2024-05-31
Pages
99
Abstract
This thesis examines whether facial expression analysis can provide a more objective view of user experience (UX) during LEGO building. Instead of relying only on self-reports, we used OpenFace 2.0 to detect facial action units (AUs) (small, standardized facial muscle movements) while people built a LEGO set at home. Eighteen participants recorded themselves during the task and then rated their enjoyment, frustration, challenge, boredom, and excitement on a 7-point scale. We trained Random Forest regression and classification models to predict these ratings from AU data. The models performed poorly. Likely reasons include that building alone produced fewer expressive facial cues and that self-reported experiences did not vary enough to support reliable prediction. The findings suggest that facial expression analysis still holds promise for UX research, but methods need refinement. Future work could include more frequent in-task self-assessments, more sophisticated machine-learning models, and combining facial data with other biometric measures to improve accuracy.
Dette speciale undersøger, om analyse af ansigtsudtryk kan give et mere objektivt indblik i brugeroplevelsen (UX) under LEGO-bygning. I stedet for kun at bruge selvrapporter anvendte vi OpenFace 2.0 til at registrere facial action units (AUs) (små, standardiserede muskelbevægelser i ansigtet), mens personer byggede et LEGO-sæt derhjemme. 18 deltagere optog sig selv under opgaven og vurderede efterfølgende deres glæde, frustration, udfordring, kedsomhed og spænding på en 7-punktsskala. Vi trænede Random Forest-baserede regressions- og klassifikationsmodeller til at forudsige disse vurderinger ud fra AU-data. Modellerne klarede sig dårligt. Mulige forklaringer er, at det at bygge alene gav færre tydelige ansigtsudtryk, og at deltagernes selvrapporter ikke varierede nok til at understøtte sikre forudsigelser. Resultaterne peger på, at ansigtsudtryksanalyse rummer potentiale i UX-forskning, men at metoderne skal finjusteres. Fremtidige studier kan omfatte hyppigere selvvurderinger undervejs, mere avancerede maskinlæringsmodeller og kombination med andre biometriske mål for at forbedre nøjagtigheden.
[This apstract has been rewritten with the help of AI based on the project's original abstract]