Computational intelligence techniques, pattern recognition and machine learning in audiovisual and biometric data analysis

Triantafyllou, Andreas M.; Τριανταφύλλου, Ανδρέας Μ.

Doctoral Thesis

Author

Triantafyllou, Andreas M.

Τριανταφύλλου, Ανδρέας Μ.

Date

2023-10

Advisor

Tsihrintzis, George
Τσιχριντζής, Γεώργιος

Keywords

Emotion detection ; Group emotion detection ; Emotional state detection ; Group concentration ; Pattern recognition ; Face recognition ; Face detection ; Tutoring system ; Αναγνώριση συναισθημάτων ; Αναγνώριση συναισθημάτων ομάδων ; Αναγνώριση συναισθηματικών καταστάσεων ; Συγκέντρωση ομάδων ; Αναγνώριση προτύπων ; Αναγνώριση προσώπων ; Ανίχνευση προσώπου ; Σύστημα απομακρυσμένης διδασκαλίας

Abstract

The analysis of the visual representation of people leads us to useful conclusions in relation to their basic external characteristics, such as the sex, skin color, etc. Furthermore, specific facial grimaces, contractions, and expansions of facial features, but also sequences of facial movements, give us useful information about the emotional state of each person. Understandably, some results can give us false information in specific cases, but this error rate can be reduced enough to cases that emotion detection is carried out on people participating in some common activities – events. In such cases, the drawn conclusions about the emotional state of each participant, give us useful information which can contribute to improving the quality of each such event. As important as it is the emotion detection of individuals, it is equally important to detect emotional states of groups of people based on common activities and events, because the “Group Affects Recognition” and the final conclusions. Specifically, in this research we have investigated, through a series of important works, the “detection of emotions of groups of people using software”, and all this research is adapted to e-learning and to tutoring systems. This research aimed to draw useful conclusions in e-teachings similar to those that experienced professors understanding in real classrooms with the physical presence of students, and accordingly adapting their lessons. For example, professors can detect situations in which students begin to get “tired” of the lesson or even bored of it. They try to adapt their lessons in order to achieve the best results in terms of imparting knowledge, which is quite efficient and “profitable” for both sides (Students – Professors), improving the quality of teaching. Thus, we managed to build an integrated system consisting of all distinct components of work, which give us real-time information about the concentration value of each group of people attending an online course. Finally, it can be used in all e-learning applications which give us the facial data of participants, because we can take the samples directly from screen output used for online teaching.