LAFTER: a real-time face and lips tracker with facial expression recognition

Abstract

This paper describes an active-camera real-time system for tracking, shape description, and classi"cation of the human face and mouth expressions using only a PC or equivalent computer. The system is based on use of 2-D blob features, which are spatially compact clusters of pixels that are similar in terms of low-level image properties. Patterns of behavior (e.g., facial expressions and head movements) can be classi"ed in real-time using hidden Markov models (HMMs). The system has been tested on hundreds of users and has demonstrated extremely reliable and accurate performance. Typical facial expression classi"cation accuracies are near 100%.

Facial Expression Recognition Based on Dimension Model Using Sparse Coding

“…Ekman considers six basic emotions: happiness, surprise, fear, anger, disgust, sadness; and categorizes facial expressions with these six basic emotions. Most research on facial expression recognition includes studies using the basic emotions of Ekman [2,3,4,5], therefore these studies have limitations for recognition of natural facial expressions which consist of several other emotions and many combinations of emotions. Here we describe research extended on the dimension model of internal states for recognizing not only facial expressions of basic emotions but also expressions of various emotions.…”

Face verification for mobile personal device

“…Typical color spaces are RGB (red -green -blue) [71], HSI (hue -saturation -intensity) [95], YIQ (luma -chrominance) [34], YCbCr (luma -chroma blue -chorma red) [181], etc. Figure 2.1: Per-pixel skin classification, blob growing, and detected face [118].…”

“…For example, histograms or charts are used in [22] [152], unimodal or multimodal Gaussian distributions are used in [127] [67]. The color models can be learned once for all, or in an online-updating manner [118].…”

“…By scanning the input image and applying pixel classification, a skin map is generated. In the next step, the skin pixels are grouped together using blob growing techniques to determine the face region [118] [67]. Fig.…”

Gestural Interfaces for Hearing-Impaired Communication

“…One of the main approaches is optical flow analysis from facial actions (Yacoob and Davis 1996;Black and Yacoob 1997;Essa and Pentland 1997;]: These methods focus on the analysis of facial actions where optical flow is used to either model muscle activities or to estimate the displacements of feature points. A second approach is using model-based approaches (Zhang et al 1998;Gao et al 2003;Oliver et al 2000;Abboud et al 2004): Some of these methods apply an image warping process to map face images into a geometrical model. Others realize a local analysis where spatially localized kernels are employed to filter the extracted facial features.…”