| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document

Title:Razpoznavanje čustvenih izrazov osebe iz slikovnega materiala z algoritmom diferencialne evolucije za izbiro značilnic
Authors:Mlakar, Uroš (Author)
Brest, Janez (Mentor) More about this mentor... New window
Files:.pdf DOK_Mlakar_Uros_2019.pdf (1,78 MB)
 
Language:Slovenian
Work type:Doctoral dissertation (mb31)
Typology:2.08 - Doctoral Dissertation
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:V disertaciji se ukvarjamo z razvojem učinkovitega programskega sistema za izbiro značilnic, na primeru aplikacije prepoznavanja čustvenih izrazov. Predlagan sistem, ki prepoznava sedem prototipnih čustvenih izrazov, vključno z nevtralnim izrazom, temelji na histogramih usmerjenih gradientov (HOG) in vektorjih razlik. Izbiro obraznih značilnic smo izvedli z uporabo ustrezno prilagojenega algoritma diferencialne evolucije za večkriterijsko optimizacijo, ki je hkrati minimiziral velikost izbrane podmnožice značilnic in maksimiziral natančnost razpoznavanja čustvenih izrazov. Razvili smo dve strategiji izbire značilnic, poimenovani "specifična” in ”splošna”. Statistični Friedmanov test je pokazal, da je ”splošna” strategija izbire značilnic primernejša. Implementiran sistem za razpoznavo čustvenih izrazov smo preizkusilina treh pogosto uporabljenih javnih podatkovnih bazah. Na podatkovni bazi Cohn-Kanade smo dosegli 98,37 % povprečno uspešnost prepoznavanja čustvenih izrazov, na podatkovni bazi JAFFE 92,75 % uspešnost in na najzahtevnejši podatkovni bazi MMI s spontanimi čustvenimi izrazi 84,07 % uspešnost. Število uporabljenih značilnic smo uspeli zmanjšati za 89 % originalne velikosti vektorja značilnic. Predlagan algoritem po uspešnosti sodi v sam vrh algoritmov za prepoznavanje čustvenih izrazov oseb, hkrati pa signifikantno zmanjša število uporabljenih značilnic, kar posledično pomeni nižjo računsko zahtevnost učenja klasifikatorjev. S to disertacijo smo demonstrirali učinkovito uporabo algoritma diferencialne evolucije za večkriterijsko optimizacijo na problemu prepoznavanja čustvenih izrazov.
Keywords:razpoznavanje čustvenih izrazov, izbira značilnic, diferencialna evolucija, razlike vektorjev značilnic, večkriterijska optimizacija
Year of publishing:2019
Publisher:[U. Mlakar]
Source:Maribor
UDC:004.932(043.3)
COBISS_ID:22195478 Link is opened in a new window
NUK URN:URN:SI:UM:DK:SDOIWYHZ
License:CC BY-SA 4.0
This work is available under this license: Creative Commons Attribution Share Alike 4.0 International
Views:362
Downloads:94
Metadata:XML RDF-CHPDL DC-XML DC-RDF
Categories:KTFMB - FERI
:
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:AddThis
AddThis uses cookies that require your consent. Edit consent...

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:English
Title:Facial expression recognition of persons from images using a differential evolution algorithm for feature selection
Abstract:This dissertation proposes an efficient feature selection system, applied to a Facial Expression Recognition system. The proposed system, which recognizes seven prototypical facial expressions, including the neutral expression, is based on the histograms of oriented gradients descriptor and vectors of differences. The selection of facial features was performed using an appropriately adapted differential algorithm for multi-objective optimization, which simultaneously minimized the size of the selected subset of features and maximized the accuracy of the facial expressions recognition. Two feature selection strategies were developed, named "specific" and "more-discriminative". The statistical Friedman test showed that the "more-discriminative" feature selection strategy is more appropriate. The implemented system for recognizing facial expressions was tested on three commonly used public databases. On the Cohn-Kanade database, an average accuracy of 98,37% was achieved, on the JAFFE database a 92,75% accuracy and on the most demanding MMI database, which is comprised of spontaneous emotional expressions, a 84,07% average accuracy was recorded. The number of features used was reduced by 89% of the original size of the feature vector. Based on the obtained results, the proposed algorithm falls into the very top of the algorithms for recognizing the facial expressions of persons, while significantly reducing the number of features used, which in turn implies a lower computational complexity of training the classifier. With this dissertation, we have demonstrated the effective use of the differential evolution algorithm for multi-objective optimization on the problem of recognizing facial expressions.
Keywords:facial expression recognition, feature selection, differential evolution, feature vector differences, multi-objective optimization


Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica