MASTER THESIS ASSIGNMENT: MULTIMODAL STRESS DETECTION FOR COLLABORATIVE TASKS WITH EARABLES
INTRODUCTION
Worker well-being affects safety and performance. Multimodal earable sensing (IMU, PPG/HR/HRV, in-ear pressure) enables path to real-time stress assessment during collaborative workshop assembly tasks.
OBJECTIVES
- Build multimodal stress classifiers and analyze the effect of stress on task performance and errors.
- Compare unimodal vs. multimodal models; assess generalization and personalization.
PROJECT DESCRIPTION
1. Literature Review: Stress assessment with wearables; validation with sensor data and/or surveys, existing datasets on stress assessment with wearables
2. Data Collection: Participation in ongoing data collection with a PhD student, in the CUBE Manufacturing facility at UT.
3. Modeling: Feature engineering + DL (CNN/LSTM/Transformer); fusion strategies (early/late).
4. Analysis: Model performance, modality importance, stress–performance correlations.
PRE-REQUISITES
Python/ML, statistics, basic experimental design; interest in HCI/IoT.
WORK
30% Theory, 50% Programming/Experimentation, 20% Writing
CONTACT
Egemen İşgüder (egemen.isguder@utwente.nl)
Özlem Durmaz İncel (ozlem.durmaz@utwente.nl)
Rob H. Bemthuis (r.h.bemthuis@utwente.nl)