Artificial odor-recognition system using neural network for estimating sensory quantities of blended fragrance

We have been studying a sensing system using a quartz-resonator sensor array and a neural network. In particular, we proposed a neural network called 'fuzzy learning vector quantization (FLVQ)' and applied it to the identification of odors and estimation of the sensory quantity that could be obtained by the human sensory test. The purpose of the present study is to construct an artificial odor-recognition system for estimating odor sensory quantities of blended fragrance, by modifying FLVQ to improve the capability of the estimation of unknown mixture compositions. It is found that the modified FLVQ has a higher capability than those of the original one and the back-propagation (BP) method in estimating sensory quantities of blended fragrances, particularly of unknown mixture composition.