	LOGIN
	>> Home
	>> Topics
	>> Students
	>> Partners
	>> Statistics

Information for topics

Topic Id:

ID topic:	579
Partner Email:	dtdim@iinf.bas.bg
Project Title:	Recognition of Basic Emotions in a Face Video Stream
Abstract:	The achievement of satisfactory results in emotions recognition could significantly improve man-machine interaction, as well as human interrelations, being an assisting technology for handicapped people. At this stage a lot of investigations in the area of emotions recognition could be found, but respective software is still hardly available. Six universal basic emotions may be defined according to literature – joy, grief, surprise, fear, disgust and anger. The neutral status can be also added to these basic emotions. A system of emotions encoding with the help of Action Units (AU) is most often applied in the research of visual emotions recognition. AU describe the movement of specific attributes of human face, e.g. lips, eyes, etc.. It is considered that each one of the basic emotions corresponds to an area in AU space. One approach to extract AU from a two-dimensional video stream is the creation of a 3D model of the face areas around specific points – mouth corners, eyes, brows and nose. The specific points could be defined either automatically or by an operator. Purpose of the diploma work offered: To apply and probably improve the achievements in the area of modeling of surfaces in video data and emotions recognition on the basis of the movement of some characteristic points on a human face with the aim to develop a free used system for emotions recognition in a video stream. The development includes the following parts: - A module for JPEG images extraction from video stream. This functionality will be provided by an external module (not included in this work); - A module for localization of the characteristic points on the face. This functionality will be: (1) supplied by an external module for automatic localization (not included in this diploma work), or (2) interactively realized by an operator; - A module for construction of a 3D model of the areas on the human face with an accent on the specific points; - A module for extracting of the action units of the specific parts on the basis of 3D model deformations; - A module for classification – by Bayes classifier or any of its known modifications; - User interface design with the possibility to open a video stream and recognition of the most probable emotion or refusal from recognition. The project will also cover the use of existing data bases of human expressions records, in order to adjust and check the accuracy of the software performance. The project will be described in its theoretic and applied aspect in the thesis. Recommended development environment: C/C++ under Windows or Linux.
Advisor:	Assoc. Prof. Dr Dimo T. Dimov
Link:
Degree:	Bachelor
Keywords:
Computer vision
Image processing