AAAI Publications, Twenty-Third International FLAIRS Conference

Font Size: 
Imitating Personalized Expressions in an Avatar through Machine Learning
Cassondra Puklavage, Alexander R. Pirela, Avelino J. Gonzalez, Michael Georgiopoulos

Last modified: 2010-05-06


This paper describes a method for automatically reproducing personalized facial expressions in a computer-generated avatar and a proof-of-concept test is implemented to evaluate the preliminary viability of the method. The method employed is a combination of an expression recognition system derived from the Facial Action Coding System, mapped by a Particle Swarm Optimization algorithm onto a computerized facial rig with ‘musculature’ based on the Facial Action Coding System. This approach is analyzed in depth, and an initial test of the algorithm is implemented and tested with promising initial results.


Computer graphics; machine learning

Full Text: PDF