SELF-ORGANIZING MULTISENSOR SYSTEMS FOR ODOR CLASSIFICATION - INTERNAL CATEGORIZATION, ADAPTATION AND DRIFT REJECTION

A system for odour classification is introduced, which is composed of both a sensor array for gas sensing and a self-organizing artificial neural network (SOM). A new classification method is proposed, which generalizes the potential function method to neural implementation in an unsupervised environment. Time-dependent analysis and drift effects are also taken into account. Results are given for a two-odour recognition experiment.