TITLE: Sensorless Human Behaviour Understanding using Computer Vision techniques
SPEAKER: Irene Kotsia (Middlesex University)

ABSTRACT:

Human behaviour analysis can be regarded as the fusion of emotion information extracted by either facial expressions or body actions. This talk will address both of these  fields. More precisely, methods that attempt facial expressions recognition by exploiting the geometrical (shape) and texture information extracted from static images or videos will be first presented. Consequently, the problem of recognizing human actions will be presented in detail. In the end, these two fields will be merged to create a novel sensorless human behaviour analysis scenario that employs the Kinect sensor to extract all available sources of information (2D and 3D). This scenario, based only on computer vision techniques to extract information, will be applied at games, thus creating a novel affective gaming scenario, in which the behaviour of the social group (consisting of the players) will be studied. Our aim is to understand the behaviour of such a social group, that  includes the recognition of the emotional state of each player, of the underlying relationships formed within the group and of their dynamics in time.