Σχεδιασμός και υλοποίηση πολυτροπικού πληροφοριακού συστήματος για την ανίχνευση γνωστικού φορτίου και ψυχολογικής επιβάρυνσης σε περιβάλλον εργασίας υπολογιστή

Μόφορης, Δημήτριος

Design and implementation of a multimodal information system for cognitive load and psychological stress detection in computer-based work environments

Master Thesis

Author

Μόφορης, Δημήτριος

Date

2026-09

Abstract

This thesis presents the design and implementation of a multimodal information system for the detection of cognitive load and psychological stress in computer-based work environments. The system integrates three non-invasive data streams: physiological signals extracted from a standard webcam via remote photoplethysmography (rPPG), continuous facial analysis for the estimation of eye strain indicators and emotional state in the Valence-Arousal space, and desktop activity metadata collected through a Flutter/Dart application using Windows APIs. The architecture adheres to a modular design principle, with each subsystem operating independently and communicating exclusively through a shared SQLite database. Data fusion is implemented via a weighted late fusion mechanism, producing a unified stress index (stress_index ∈ [0,1]) per five-minute window. Scoring is individually calibrated, computed as deviation from the user's personal resting baseline rather than general population-level thresholds, thus addressing a central limitation identified in the existing literature. The evaluation was conducted at a technical and functional level through automated tests and functional usage scenarios. Additionally, a comparative evaluation of two locally executed language models (Llama 3.1 8B and Qwen3.5:4b) was performed using the DeepEval framework with an independent judge model (Gemma 4:12b). The absence of an extended pilot study with participants is acknowledged as a limitation, particularly with respect to user acceptance of interventions and perceived privacy. Nevertheless, the technical implementation demonstrates that a multimodal approach combining personalised baseline calibration and fully local processing constitutes a feasible direction for well-being support systems in computer-based work environments.

Postgraduate Studies Programme

Πληροφοριακά Συστήματα και Υπηρεσίες

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών Συστημάτων

Number of pages

138

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/19822

Collections

Τμήμα Ψηφιακών Συστημάτων

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα