Facial Feature Analysis of Spontaneous Facial Expression

The quick advances in key computing technologies, the heightened user expectation of computer interfaces, and the ever-increasing complexity of applications, have made the development of emotionally intelligent interfaces both necessary and possible. Indeed, the true strength of facial expression recognition shows when seamlessly integrated into mainstream interfaces enabling applications to add facial expressions to traditional keyboard and pointing input modalities. Spontaneous facial expression occurs more frequently than its disjoint prototypic counterpart in everyday interaction. However its automated analysis continues to pose a number of challenges. In this paper we address the dynamics of spontaneous facial expression. We build a model for natural facial expression and a feature point tracking methodology that tracks spontaneous expression over time. The methodology improves the efficiency of optical flow based tracking by utilizing motion characteristics of individual feature points across frames of an image sequence. The Non-exhaustive Sum of Squared Differences (SSD) is developed as an optimisation on the SSD algorithm on which many tracking methodologies are based. For comparable tracking accuracy, the presented feature point tracking methodology yields a tenfold increase in efficiency when compared to existing optical flow methodologies.