Abstract
Existing smell classification devices can display the strength of smells as quantitative values by using electric circuits. But they cannot perform the type of smells such that they are good or bad. In this paper, we construct a smell measurement system that can classify the type of smell by using neural networks and metal oxide semiconductor gas sensor arrays. In particular, we classify the kind of wines, which has been considered as the most difficult classification. We measure the smell data of white wine and red wine by using a method to change temperatures of smells of wine data and classify them by using a neural network. Using the classification rate of white wine and red wine, we consider the ability of classification of the proposed method.