Skip to Main content Skip to Navigation
Theses

Plate-forme multimodale pour la reconnaissance d'émotions via l'analyse de signaux physiologiques : Application à la simulation d'entretiens d'embauche

Abstract : Emotion recognition of is an important aspect of affective computing which the main objective is the study and the development of emotional and behavioral interactions between human beings and interactive conversational agents. In this context, it is important to point out the use of both input devices and tools for signal processing, which lead us to the estimation of the user's emotional state. The work presented in this manuscript describes the development of a multimodal platform for the acquisition and the processing of physiological signals. This platform can be considered as a modular and generic middleware allowing to measure, in real time, the emotional state of the user via the estimation of a state vector. The platform is being developed in the context of the PISE project that aims at the development of a job interviews simulator. This simulator allows students or job seekers to train to master their emotional and behavioral skills. The proposed approach is based on experimental results and associated with an original methodology involving different mathematical models. Different experimental protocols based on two techniques of induction of stimuli (IAPS pictures and video clips of Schaeffer) where proposed. They allow the determination of the correlation between the emotional states and the physiological signals (EEG, ECG, etc.). Three classification methods (SVM, naive Bayes, logistic regression) were used and compared on several criteria of evaluation. We proposed, from the experimental results, a methodology allowing the real-time extraction of the emotions via the proposed models. An experimental study was conducted with the aim to validate the multimodal platform through the recognition of emotional states from video sequences developed at the University of Angers. The proposed multimodal approach gave better results than the previous uni-modal approaches. Finally, our platform was integrated into the job interview simulator and evaluated through both subjective and objective collected data. The results allowed to partially validating our approach and the simulator.
Document type :
Theses
Complete list of metadata

Cited literature [220 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00997249
Contributor : Anne-Marie Plé Connect in order to contact the contributor
Submitted on : Tuesday, May 27, 2014 - 5:02:46 PM
Last modification on : Wednesday, October 20, 2021 - 3:19:29 AM
Long-term archiving on: : Wednesday, August 27, 2014 - 12:06:00 PM

Identifiers

  • HAL Id : tel-00997249, version 1

Citation

Hamza Hamdi. Plate-forme multimodale pour la reconnaissance d'émotions via l'analyse de signaux physiologiques : Application à la simulation d'entretiens d'embauche. Modélisation et simulation. Université d'Angers, 2012. Français. ⟨tel-00997249⟩

Share

Metrics

Record views

1311

Files downloads

4459