The machine handling of complex sensory stimuli is an extremely important but, as yet, elusive engineering goal. Biological organisms are good at detecting and identifying 'out-of-the-ordinary' events. Machines function best in unchanging, stationary environments, in which all event types are represented in the training data. The goal of the project is to address this machine weakness.
We will design and develop an environment-adaptive autonomous active system that will detect and identify rare events from the information derived by multiple, active information-seeking sensors. Such a system will probe for relevant cues, will autonomously adapt to new and changing environments, and will reliably discard non-informative data.
The project offers an integrated approach, ranging from neurophysiological research at the single cell level, up to building an engineering system. We aim at safety and security applications, but the results of the development will be equally relevant to a wide range of other applications in information and communication technology, unobtrusive monitoring in healthcare, and intelligent devices that enhance human cognitive performance.
The progress in our research will be measured by two criteria:
- qualitative - consistency of new models with physiological and psychophysical data, and
- quantitative - including the newly developed techniques in state-of-the-art information processing technology and evaluating the improvements in performance.
To achieve our goals, we have assembled partners with a strong motivation to advance the basic science of understanding and modeling cognitive systems, with equally strong interest in building useful technology.
Besides the fulfilment of our scientific and technological goals, another important result of DIRAC will be long lasting collaboration between cognitive sciences and engineering, resulting in a new generation of researchers, trained in both cognitive disciplines and in engineering.