The performance of any electronic nose is ultimately determined by the properties of its constituent parts (e.g., the sensors, signal processing and pattern-recognition engine). Electronic noses currently exploit the technologies of several classes of sensor material (e.g. semiconducting oxide, conducting polymer, phthalocyanines and lipid coatings) as well as a variety of pattern-recognition paradigms (e.g. back-propagation, self-organizing map and discriminant function analysis). Consequently, there is a need to compare objectively the performance of the increasing number of both research and commercial electronic noses. This paper addresses this problem and suggests the need for odour standards to quantify both the ability of an electronic nose to discriminate between similar odours (i.e. its 'resolving power') and a number of dissimilar odours (i.e. its 'range'). We present a generic model from which we can define these two fundamental parameters, and hence develop a benchmark for the performance of these different electronic noses against two proposed odour standards. This model can be employed not only as a design tool to predict the performance of an electronic nose against an odour standard, but also as a diagnostic tool that can determine, for example, the effect of random errors in the sampling method, sensor characteristics or the effect of systematic errors associated with sensor drift or changes in ambient temperature. We believe that our definition of odour standards and performance parameters for electronic noses could be used to create a European standard, which is now required in this rapidly expanding field and marketplace.
