Search results

Query: faculty: "FNWI" and publication year: "2009"

AuthorBruno Jakic
TitleHuman Facial Features Extraction and Classification using Domain-Sensitive HMMS : A component for practical application of Nomadic Information
SupervisorFrank Nack
Year2009
Pages29
FacultyFaculty of Science
ProgrammeFNWI BSc Kunstmatige Intelligentie
AbstractRecent advances in computing power have brought the potential for real-time applications of image and facial recognition from the laboratory to the every-day reality. One of the possible uses of these methods is in the field of Nomadic Information, allowing for generation of media feedback (e.g. displaying advertisements) based on subject’s individual or general profiles as observed through an imaging system. While different methods of extracting the features that make up the profiles exist, most of them are extremely specific (e.g. “gender” only) or not very robust. This paper presents a generalized, robust and scalable method of facial feature extraction based on two-dimensional Hidden Markov Models [2], extended by addition of a domain context to the model. To validate this method, classification results are compared with the results obtained by specialized classifiers from literature under controlled and uncontrolled conditions.
Document type scriptie bachelor
Download paper