10:00   Oral Session 7-GZ – Facial Expression Recognition
Chair: Kostas Karpouzis
10:00
25 mins 		Evaluating AAM Fitting Methods for Facial Expression Recognition
Akshay Asthana, Jason Mora Saragih, Michael Wagner, Roland Goecke
Abstract: The human face is a rich source of information for the viewer and facial expressions are a major component in judging a person's affective state, intention and personality. Facial expressions are an important part of human-human interaction and have the potential to play an equally important part in human-computer interaction. This paper evaluates various Active Appearance Model (AAM) fitting methods, including both the original formulation as well as several state-of-the-art methods, for the task of automatic facial expression recognition. The AAM is a powerful statistical model for modelling and registering deformable objects. The results of the fitting process are used in a facial expression recognition task using a region-based intermediate representation related to Action Units, with the expression classification task realised using a support vector machine. Experiments are performed for both person-dependent and person-independent setups. Overall, the best facial expression recognition results were obtained by using the Iterative Error Bound Minimisation method, which consistently resulted in accurate face model alignment and facial expression recognition even when the initial face detection used to initialise the fitting procedure was poor.
10:25
25 mins 		Get The FACS Fast: Automated FACS face analysis benefits from the addition of velocity
Timothy R Brick, Michael D Hunter, Jeffrey F Cohn
Abstract: Much progress has been made in automated facial image analysis, yet current approaches still lag behind what is possible using manual labeling of facial actions. While many factors may contribute, a key one may be the limited attention to dynamics of facial action. Most approaches classify frames in terms of either displacement from a neutral, mean face or, less frequently, displacement between successive frames (i.e. velocity). In the current paper, we evaluated the hypothesis that attention to dynamics can boost recognition rates. Using the well-known Cohn-Kanade database and support vector machines, adding velocity and acceleration decreased the number of incorrectly classified results by 14.2% and 11.2%, respectively. Average classification accuracy for the displacement and velocity classifier system across all classifiers was 90.2%. Findings were replicated using linear discriminant analysis, and found a mean decrease of 16.4% in incorrect classifications across classifiers. These findings suggest that information about the dynamics of a movement, that is, the velocity and to a lesser extent the acceleration of a change, can helpfully inform classification of facial expressions.
10:50
25 mins 		Dynamic Cascades with Bidirectional Bootstrapping for Spontaneous Facial Action Unit Detection
Yunfeng Zhu, Fernando De la Torre, Jeffrey Cohn, Yujin Zhang
Abstract: Automatic facial action unit recognition from video has been a long standing problem in facial expression analysis. Several issues contribute to the challenge of this task. These include non-frontal pose, moderate to large out-ofplane head motion, large variability in the temporal scale of facial gestures, and the exponential nature of possible facial action combinations. Most popular methods pose the AU recognition as a binary classification problem using different features (e.g. appearance, shape) and classifiers (e.g. Boosting, support vector machines). Although existing systems have reported improved results, an important and relative unexplored problem in AU learning-based recognition systems is how to select the positive and negative samples optimal for AU recognition. Typically, the peak frames of the AUs are selected as positive class and the negative samples are randomly selected from other AUs. However, this approach might suffer from two main drawbacks: (1) if there is a large number of training samples, some classifiers such as Support Vector Machines (SVMs) do not scale well with the number of samples (e.g. O(n3) for the worse case in SVM), (2) given the manual FACS coding, it is not clear how to choose the positive and negative samples. If we only label the peaks as positive samples, there will be a large imbalance between positive and negative samples, especially for infrequent AU. On the other hand, if all the frames from onset to offset are labeled as positive, the number of false positive samples will increase. In this paper, we propose the Dynamic Cascades with Bidirectional Bootstrapping method to automatically select the positive and negative class samples in training. The results reveal improved results in the recognition performance of important AUs on the RU-FACS Spontaneous Expression Database.