Facial component-based blended facial expressions generation from static neutral face images

Abstract

Facial expression synthesis is getting a wide-spread attention since past several years due to its multimedia applications. In most of the earlier research works, example images of target expressions are required to produce synthesized facial expressions. The paper aims to generate six basic and twelve blended facial expressions from a static and RGB neutral face image without any exemplar of expressive face images. The proposed automatic expression generation system consists of several sub-systems, namely, a knowledge-based system, a module for symbolic formulations of basic and blended facial expressions, an expressive facial components generator and an expressive face generator. The knowledge-based system stores the normalized facial feature parameter values. Symbolic formulations of facial expressions are used to reconstruct facial expressions from a static neutral face image using the parameters stored in the knowledge base. Expressive facial components generator performs automatic expressive facial feature generation as well as automatic facial feature extraction and landmark annotation. Finally, expressive facial components are combined to produce an expressive face in expressive face generator. The system generated expressive face images are validated in four different ways: inter-rater reliability measure, similarity measurement in the frequency domain, similarity measurement using SSIM, FSIM, HOG features and accuracy measurement using both appearance-based and geometry-based feature extraction methods. The geometry based feature extraction method generates 90% recognition accuracy for system generated face images.