Smell sensing system was proposed by MORIIZUMI et al. and refers to a computer system that carries out the measurement, learning and discrimination of various kinds of odors. Newly-developed smell sensing system employs ON/OFF switching control of the heaters. As for the discrimination of odors, Variable Interaction Method, an original method which is based upon Bayesian discriminant algorithms was proposed.

In the process of this research, it was found that proposed method and neural network share an interdependent nature. By taking advantage of the nature of both methods the smell discrimination was performed. In addition, suitability of gas detectors for selected odor samples was estimated as well as the selection of odor samples for gas sensors. The effectiveness of the method in this research is demonstrated by the results of experiments.