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

Information for topics

Topic Id:

ID topic:	473
Partner Email:	L.J.M.Rothkrantz@tudelft.nl
Project Title:	FED: An online facial expression dictionary as a first step in creating a nonverbal dictionary
Abstract:	A verbal dictionary can be used to look up the spelling of a word, sometimes the phoneme representation, the meaning in different contexts and rules of transformation. The goal of this graduation project was to develop a prototype of an online Facial Expression Dictionary, or FED for short, as a first step in the creation of a complete Nonverbal Dictionary. A complete Nonverbal Dictionary would contain information about all the ways people communicate with each other nonverbally. Instead of words, FED contains information about facial expressions. FED had to become available as a website. The first step in the creation of FED was the definition of an entry in FED. The choice was made to base each entry in FED on a facial expression picture generated by a facial expression generation tool called FaceShop. The next logical step was to implement management facilities, with which the FED entries could be managed. Subsequently, the FED database was filled with 56 facial expressions. Essential for a Nonverbal Dictionary is the possibility to issue a nonverbal query through (multimodal) content. With FED, issuing a nonverbal query is done through uploading a picture containing a facial expression, after which the user semi-automatically determines the location of the face and the coordinates of the 30 Facial Characteristics Points or FCP’s of the face model defined by Kobayashi and Hara. FED then determines the label of the unknown facial expression by comparing the FCP coordinates to the FCP coordinates of all entries present in the database. Also, it is possible to let FED determine the la bel of a facial expression sketched with FaceShop. Other query possibilities have been implemented as well. It is also possible to look for entries in FED on facial expression label, active Action Units or specific geometrical features. Finally, it is possible to look for entry incrementally, were the user iteratively selects the facial expression that resembles the facial expression he is looking for the closest. The concept of FED as an online Facial Expression Dictionary was tested and found to be a viable approach. The FED system is easy to use, adapt, extend and manage. The query possibilities have been tested by a group of 30 students, and although there is room for some improvement, the results are satisfactory. The approach taken with FED could be used to create a complete Nonverbal Dictionary.
Advisor:	Leon Rothkrantz
Link:
Degree:	Master
Keywords:
Computer Software
Algorithms & problem solving
Artificial intelligence & Neural networks
Computer vision
Data mining
Data modeling
Image processing