The SAMoVA group focuses its research activities mainly on audiovisual content. More precisely, the research works are addressing the following main topics:

  • Structuring: most of the works on that topic aim at analyzing the way the information in an audio / video signal is organized. The goal is to localize segments at different temporal scale providing pieces of information for direct access to the content, or for summarization. In the scope of this topic are also included works aiming at comparing and evaluating similarity between contents in order to organize heterogeneous databases in collections for example.
  • Analysis: the research group develops methods for new types of feature extraction or exploitation. Those features are most of the time used as inputs for upper-level recognition or structuring tasks.
  • Modeling: for recognition tasks, multimedia objects (speakers, languages, human shapes, etc) have to be modelized though a set of parameters which have to be tuned by dedicated algorithms during a learning step or on-line, during the content analysis.

Those works are applied on different kinds of content for different kinds of applications such as:

  • Structured audiovisual content analysis: the goal here is to automatically determine temporal boudaries which are intentionnaly user to structure an audiovisual recording or stream.
  • Spoken content analysis: applications in that field are dedicated to extract discriminant features from a spoken content in order to characterize different aspects of this content such as the language or the speaker.
  • Media and video surveillance: the goal here is to be able to identify specific predefined events or objects, and to offer tools helping an end-user in a task of audiovisual content comparison.
  • Mutimedia indexation engineering: that last part of the works is related in the development of technological tools which are dedicated to ease the integration of multimedia applications.