Abstract
Recently, ANNs (artificial neural networks) have been widely used in odour recognition using an artificial olfactory system, an electronic nose. However, when classes have a great difference in shape, volume and density or they are overlapped with each other, the current ANN methods sometimes fail to get a good classification. To solve these problems, an improved odour-recognition method that combines the FCMA (fuzzy c-means algorithm) with an RBF (radial basic function neural network) is described in this paper. Application of this method to the odour-pattern classification of some gas mixtures using an electronic nose based on a gas-sensing sensor array has been carried out. Both the computer simulations and the odour-recognition experimental results demonstrate the unusual effectiveness and the good recognition performance of the above-mentioned method.