ABSTRACT
In the last years, a growing interest in mimicking the mammalian olfactory system has arisen with the aim of developing a computer-based system which is able to classify odours and perhaps to measure odour intensity and nature. Such a system is called an electronic nose [I]. Typically, electronic noses integrate an array of a few sensors with partially overlapping sensitivities to odours, and a pattern recognition system [2]. Various types of sensor technologies as well as pattern recognition techniques are available. Nonlinear pattern-recognition methods have proven more successful than linear ones thanks to their ability to handle the highly non-linear transduction properties of the sensors and 'to compensate for sensor drift.
