Πλοήγηση ανά Συγγραφέας "Vazakopoulou, Kalliopi-Marina"

Τώρα δείχνει 1 - 2 of 2
  • Μικρογραφία εικόνας
    Τεκμήριο
    Categorical assessment of depression based on high level features.
    (Τ.Ε.Ι. Κρήτης, Σχολή Τεχνολογικών Εφαρμογών (Σ.Τ.Εφ), ΠΜΣ Πληροφορική και Πολυμέσα, 2018-04-17) Vazakopoulou, Kalliopi-Marina; Βαζακοπούλου, Καλλιόπη-Μαρίνα
    Depression is one of the leading causes of ill health and disability worldwide. In the US nearly 50 million people were living with depression in 2015. According to the latest estimates from the World Health Organization (WHO), more than 300 million people worldwide are living with depression, an increase of more than 18% between 2005 and 20151. This fact seems to be confirmed by an EU Green Paper [1], which manifests that one to four civilians suffer from a mental illness at some point during their lifetime, sometimes even causing suicidal tendencies. Objective measures of depressive symptomatology could be advantageous for clinicians, in the context of a clinical decision support system. Automatic recognition of emotions through intelligent systems, established to assess facial expressions is a vital issue that needs to be addressed for understanding human behavior, interpersonal relationships, and most importantly for mental health assessment. According to the literature in the field of clinical research, facial expressions were used to assess possible deficiencies or weaknesses in emotional expression as well as in social diagnosis in psychiatric-psychological disorders [2] [3]. Major Depressive Disorder (MDD) is the most common mood disorder, varying in terms of severity, impairing an individual's functionality and ability to deal with their daily routine. The figures provided by the EU Green Paper [4], [1] are alarming, as it is argued that by 2020 MDD is expected to become the main cause of disability. The aim of this dissertation is to develop a framework capable of detecting objective signs of MDD, to support the clinical care of patients. More specifically, the proposed system aims to identify signs related to MDD as portrayed in the facial expression of an individual. This has been designed within the terms of the AViD-Corpus (AVEC’14) [5], which provides audio-visual recordings annotated for the level of depression based on self-reports. The proposed framework was evaluated through experimental tests, designed to detect patterns related to depression in the facial expression. In the proposed implementation the emotions of happiness and sadness, two of the six basic emotions according to Ekman and Friesen [6], [7], are examined more carefully, as they seem to be respectively negatively and positively correlated with MDD. For example, the emotion of sadness appears to be more prominent in individuals suffering by MDD compared to that of happiness. In summary, the primary purpose of this thesis is to design and develop an application in MATLAB for facial image analysis, with the ultimate goal of detecting visual signs of depression through video recordings. As it is mentioned above, Facial expression is significant in human interaction and communication since it contains critical information regarding to emotion analysis. High level information about the facial features was extracted with the use of OpenFace toolkit [8], while special focus has been given to the AViD-Corpus (AVEC’14) for testing the developed algorithms. Subsequently, since high-level information was extracted by the facial features, the proper machine learning algorithms were selected, in order to examine the sensitivity and specificity of this proposed framework. Several classification algorithms were tested, namely: Discriminant Analysis, Random Forest Tree, Naïve Bayes, Linear and k-nearest Neighbor Classifiers. The best performing method involved the Discriminant Analysis classifier, with a runtime of approximately 4 minutes, to be precise 4,29 minutes. Additionally, by selecting an approximately 1 second-window, with Leave-One-Subject-Out cross-validation an accuracy of 72.57% was achieved for our depression assessment framework.
  • Μικρογραφία εικόνας
    Τεκμήριο
    Κρυπτογράφηση επεξεργασμένης εικόνας σε ενσωματωμένα συστήματα.
    (Τ.Ε.Ι. Κρήτης, Τεχνολογικών Εφαρμογών (Σ.Τ.Εφ), Τμήμα Μηχανικών Πληροφορικής Τ.Ε., 2013-12-19T09:01:27Z) Βαζακοπούλου, Καλλιόπη-Μαρίνα; Vazakopoulou, Kalliopi-Marina
    Τα πολύπλοκα πολυπύρηνα ενσωματωμένα συστήματα γίνονται όλο και πιο διαδομένα στα σύγχρονα προϊόντα μικροηλεκτρονικής. Η ανάπτυξη αυτών των συστημάτων, δημιουργεί νέες δυνατότητες εξέλιξης σε πολλούς τομείς της επιστήμης και της τεχνολογίας. Η ανάγκη υλοποίησης όλο και περισσότερων εφαρμογών σε τέτοια ανεπτυγμένα συστήματα, τα οποία παρέχουν αποτελεσματικότερη και πιο αξιόπιστη απόδοση αποτελεί αντικείμενο έρευνας της παρούσας εργασίας. Θέμα της παρούσας πτυχιακής εργασίας είναι, η υλοποίηση αλγορίθμων κρυπτογράφησης σε πολυπύρηνα ενσωματωμένα συστήματα. Επιπλέον εξετάζεται συγκριτικά η απόδοση τεσσάρων αλγορίθμων κρυπτογράφησης, σε ένα σύστημα δύο πυρήνων που υλοποιείται σε μια πρωτότυπη πλατφόρμα της Xilinx. Σε επόμενο στάδιο επιχειρείται η αξιολόγηση της αξιοπιστίας της κρυπτογράφησης/αποκρυπτογράφηση σε μια επεξεργασμένη εικόνα, σε σχέση με τα αποτελέσματα που προέκυψαν από έναν αλγόριθμο επεξεργασίας εικόνας του MATLAB.