A Representation Language for Multimedia Web Semantic

While the extraction of low level multimedia features has been largely studied, the automatic extraction of high level descriptions is still an open research field. Moreover, as semantic Web languages, which are based on description logics, are conceived for describing all kinds of resources, specific extensions of semantic languages for describing multimedia resources are needed. Descriptors offering enough expressivity to represent spatio-temporal constraints are the major needed tools to describe video high level features. Such descriptors should allow to bridge the different semantic levels of descriptions and to narrow the semantic gap. In this paper we present a generic language for the representation and the description of multimedia contents. Based on finite automata and conceptual graphs, this formalism allows to express spatial and temporal constraints, as to define sequences of events in a video and positions of objects in an image. In addition to the description of complex concepts, this formalism allows to monitor behaviors in a video-surveillance context, to detect important highlights in a sport context and also to validate MPEG7-based descriptions.