An artificial olfactory system using tiered artificial neural networks

A method for implementing an electronic nose, trained to discriminate between different odorants is demonstrated. The `nose´ consists of a number of chemically modified gas sensors. On exposure to odorants, a unique dynamic response pattern (termed “kinetic signature”) is obtained for each sensor, based on the interfacial kinetics of the odorant and sensor coating. These kinetic signatures are used to train a multi-tiered artificial neural network (ANN) to discriminate between different odorants. Subsequent recognition of odorants presented to the `nose´ is by means of identifying two dimensional olfactory response maps, generated from the outputs of the ANN