About this project
Humans have long been interested in duplicating our senses with artificial sensors. The sense of smell in particular has fascinated researchers as it is a sense that is relatively poorly understood yet valuable for a number of applications. In industrial processes that deal with odour monitoring, there is often high level information that consists of knowledge about a specific odour character, information about a substance´s chemical composition or information about the processes or actions which cause specific gases to be released. At the same time, gas sensing technologies are increasingly used in industrial settings, either as simple gas sensors or more complex electronic noses using an array of sensors coupled to pattern recognition algorithms. The purpose of this project is to bridge a gap which exists between the low level sensor data and the knowledge that humans have about odours, such that correlations between what is sensed and what is known can be automatically obtained. This is achieved by mining textual data into new representations which can be reasoned with, and connecting these representations to the processed sensor data. New
visualisation tools will also be developed which enable users to verify the assigned correlations. The net ambition is to redefine our notion of electronic noses from low level instruments to systems which can readily exploit high-level human knowledge in order to obtain a complete assessment of odours.
On grant from Vetenskapsrådet, Swedish Research Council (about 1,600,000 SEK).