Contact person: Nedelko Grbic
Multisensor signal processing
Multisensor signal processing has been an important topic of research for many decades. Traditionally, the main applications have been in the military field such as in radar/sonar detection and tracking of enemy transportation. Recently the use of array processing has played a major role in the cellular mobile telephone network in order to increase capacity.
It is anticipated that a significant near-future growth of acoustic array solutions will take place in the field of speech enhancement for handheld and hands free telephone equipment. While this fact remains to be seen, a speech array revolution is already under way in the field of hearing aids. This area is strongly dependent on successful speech acquisition. The problem thus far in speech array processing has been a poor correspondence between the theoretical formulations and practical circumstances which has lead to solutions that are excessively dependent on accurate sensing. The aim of our research is to analyze and develop array formulations which may appropriately handle the diversity in array signal processing applications. Another important goal of the research is to deliver a transparent link between the theoretical array framework and the physical real-time world. The most promising approach has shown to be linked to information theoretic models, assuming only independence between different sources, to separate and/or extract individual physical sources without any other a-priori information about the channels or the sources, i.e. blind methods in signal processing.