Σκοπός της παρούσας εργασίας είναι να παρουσιάσει τους τρόπους που μπορεί το στρες να ανιχνευθεί με τη χρήση εξελιγμένων αλγορίθμων Μηχανικής Μάθησης, χρησιμοποιώντας ως δεδομένα φυσιολογικές μετρήσεις του ατόμου καθώς και χαρακτηριστικά του προσώπου του.
Stress has always been a major problem for individuals and societies. But in recent years, due to the modern urban lifestyle, the problems due to stress are getting even bigger. Thus, the need for timely and reliable detection of stress in individuals to deal with any symptoms as quickly and effectively as possible becomes even more urgent. The purpose of this dissertation is to present how stress can be detected using advanced Machine Learning algorithms, using physiological measurements or facial characteristics of the individuals as input data. The measurements are usually collected traditionally, i.e. as part of an experiment in the laboratory. However, the use of wearables by a huge portion of the population in recent years facilitates the collection of physiological data and greatly increases their volume, and therefore it improves further the reliability of the Machine Learning algorithms used to detect stress. The dissertation is divided into 5 chapters. In the first, we introduce the problem and the theory of emotions. In the second we present the theoretical background around stress. In the third, we review the methods of detecting stress either through facial features or through laboratory exams and present how technology can help through the sensors of wearable devices or even the use of virtual reality technology. In the fourth, we record the Machine Learning methods that exist in the literature of the last decade for the detection of stress and in the fifth, we have a discussion and we record some conclusions mainly regarding the effectiveness of the Machine Learning methods in the detection of stress.
Επισκόπηση τεχνικών ανίχνευσης στρες μέσω χαρακτηριστικών του προσώπου, φυσιολογικών δεδομένων και Μηχανικής Μάθησης