Human being has communicated using five senses, seeing, hearing, touching, smelling, and tasting. Among them only seeing and hearing are too much evaluated and the remaining ones are not important from view point of information processing. But if we wish to grade up the information quality, the remaining senses should be considered. In this paper, we consider smell sensing and classification, which is important to grade up the information quality. We use metal-oxide semiconductor gas sensors to measure the smells and classify  the data by using neural networks. Using our approach enables us to develop new technology, for example, television with smell communication ability, health-care products to sense our breath, etc.