Summary

2011 International Symposium on Nonlinear Theory and Its Applications

2011

Session Number:A4L-E

Session:

Number:A4L-E3

Recognizing Facial Actions Using RBF Networks

Koichi Takahashi,  Masaki Nakanishi,  Yasue Mitsukura,  

pp.298-301

Publication Date:2011/9/4

Online ISSN:2188-5079

DOI:10.34385/proc.45.A4L-E3

PDF download (1.6MB)

Summary:
In this paper, we propose a novel method to express actions especially based on facial expressions for avatar’s motion control. Recently, avatar systems for web conference tools are expected to be applied to usual PCs in order to protect private information and entertain other persons. Most preexisting methods can distinguish several facial expressions, but cannot consider the intensity of the facial expressions. Then, we propose a facial expression intensity estimation method using radial basis function (RBF) networks which is able to define the partial regions corresponding to arbitrary facial expression. Since the proposed method estimates the intensity of expressions, we can display not only the basic expressions but also the deformed actions. The experimental results show that the proposed method effectively outputs the intensity of three kinds of facial expressions that are positive, negative, and neutral, and can be applied to the avatar systems.