In this paper, the effects of dynamic facial expression properties on the recognition of emotions were investigated. Although many studies of facial expression recognition and analysis have been conducted, the stimuli used in these experiments were typically static. Even when dynamic facial expressions were used, the experimental task was restricted to the basic expressions. In natural communication environments, facial expressions are often subtle or ambiguous. Studies of the analysis and synthesis of subtle and dynamic facial expressions are therefore required. As a quick going up expression could be recognized as a different emotion from a slow rising one, we represented facial expressions as the combination of time function of intensity and presentation speed. Experiments using stimuli that were combinations of four intensity time functions and three presentation speeds were conducted. Participants rated the perceived intensity of the emotions depicted, as well as how realistic they appeared to be. Time function and speed had different effects on different emotions, i.e. perceptions of anger were affected by set up speed, and the perceived intensity and realism of smiling expressions were also affected by speed.

Inspired by Hebb's cell assembly theory about how the brain worked, we have developed a function localization neural network (FLNN). The main part of a FLNN is structurally the same as an ordinary feedforward neural network, but it is considered to consist of several overlapping modules, which are switched according to input patterns. A FLNN constructed in this way has been shown to have better representation ability than an ordinary neural network. However, BP training algorithm for such FLNN is very easy to get stuck at a local minimum. In this paper, we mainly discuss the methods for improving BP training of the FLNN by utilizing the structural property of the network. Two methods are proposed. Numerical simulations are used to show the effectiveness of the improved BP training methods.

Use of uncalibrated images has found many applications such as image synthesis. However, it is not easy to specify the desired position of the new image in projective or affine space. This paper proposes to recover Euclidean structure from uncalibrated images using domain knowledge such as distances and angles. The knowledge we have is usually about an object category, but not very precise for the particular object being considered. The variation (fuzziness) is modeled as a Gaussian variable. Six types of common knowledge are formulated. Once we have an Euclidean description, the task to specify the desired position in Euclidean space becomes trivial. The proposed technique is then applied to synthesis of new facial images. A number of difficulties existing in image synthesis are identified and solved. For example, we propose to use edge points to deal with occlusion.