PATTERN-RECOGNITION IN GAS-SENSING - WELL-STATED TECHNIQUES AND ADVANCES

A conceptual systematization of pattern-recognition techniques that are or could be applied to the gas-sensing field is provided. In particular, the Bayesian method and artificial neural networks are considered. Inner similarities, i.e., extended use of discriminants, drawbacks and possible improvements are also reported. The context of artificial olfaction has also been taken into account, in particular the physiological aspects, as an important area where enhanced pattern-recognition techniques can be fruitfully applied. In this respect 'fuzzy set theories' and 'self-organizing maps' are also illustrated as promising new tools for increasing the overall performance of pattern recognition.