Cognitive Systems

Key aspects of research

The ever advancing pervasion of everyday life with diverse technical assistants calls for intelligent systems capable of interacting with their surroundings in a natural and human- like way. This requires technical systems to understand their environment, i.e., they have to create a semantic representation they can analyze to asses and predict the current situation. Main issues herein are uncertain, distributed and multimodal perception, extracting relevant information, and projecting it to a symbolic level, as well as scene interpretation.

Methodology

The main sources or transmitters of information about a current scene for humans are acoustic and visual signals. To extract relevant features from this data, probabilistic signal processing at different levels of abstraction is performed. Methods for object detection, classification, and tracking are used to create a semantic representation of the human- centered environment.

Technical equipment

  • Laboratory for visual and acoustical perception
  • Microphone arrays for acoustic source localization