Symbolically Describing the Sensory Perceptions from an Artificial Nose

For the past two decades, electronic noses are being developed to serve a wide variety of industrial applications in addition to providing many insights into biological systems. A large part of the development into olfaction has been focussed on the sensor technology and development of sensor systems capable of measuring chemical gases [3]. To complete the model however, consideration needs also be given to the interpretation of the sensor data and the communication of these interpretations to an end user. This is especially important in applications where human and sensor systems need to cooperate in assessing the quality or the nature of an odour. It is also useful in the case where the human user is not an expert in artificial sensor systems for example in applications such as quality control in the food industry, detection of hazardous gases and evaluation of odour characteristics. In order for electronic noses to be useful in many applications, there is a need to correlate the human conception of odours to the electronic one. One way to satisfy this need is to ground a common language to the perceptions of both sensor systems. Determining how to appropriately label olfactory signals is a challenging problem. This challenge is due mostly by the fact that for humans there is no direct correlation between the name given to odours and the chemical composition of an odour. Common odours are often described by their quality or the experience felt when sensing particular smells and are rarely identified in terms of their chemical nomenclature or fixed standards [1]. Therefore facing the possibility that the progressing sensor technology in electronic noses would in many ways be able to outperform humans in the detection of odours, it is highly advantageous for such a system to be able to convey its own perceptions on a symbolic level. The main challenge, however, is to correctly represent the sensor data with a set of predefined labels. The labels then serve as a common lexicon between the human and the electronic device. Additionally, the system would need to be able to manipulate the symbols to accord with its sensor perceptions. Therefore, the system maintains the freedom to evaluate the appropriateness of the given symbols with respect to the sensory analysis space, thus possessing the capability to communicate or describe its own perception of odours using a common lexicon. To cluster the data from the electronic …