Odor classification by neural networks

It is important to detect an odor in the human living space and artificial electronic noses have been developed. This paper considers an array sensing system of odors and adopts a layered neural network for classification. We use all measurement data obtained from fourteen metal oxide semiconductor gas (MOG) sensors. Some sensors are not sensitive while others are sensitive. In order to classify odors, we use data from all fourteen sensors even if some of them are not sensitive so much. We will propose three methods to use the data by insensitive sensors to find the features of odors. Then, applying those features to a layered neural network, we will compare the classification results.