Realization of recording a wide range of odor by utilizing both of transient and steady-state sensor responses in recording process

Abstract

An odor recorder is a system to record and reproduce odors. In order to record wide a range of odors, a number of available odor components are required in an odor recorder. Although many odor components were available in a previously developed odor recorder using an automatic sampler, there is still a collinearity problem caused by the lack of pattern separation among odor components. Then, we propose a new odor recorder with a wider recording range based on both transient and steady-state sensor responses in recording process. In this system, time constants are extracted from QCM sensor responses using the auto regressive (AR) model and are used together with conventional steady-state frequency shifts as variables to discriminate among odor components. Based on this proposed method, the pattern separation among odor components can be improved and up to 13 odor components can be used with a minimized collinearity problem. Recipes of five target odors with different impressions were successfully estimated based on the same set of these 13 odor components and sufficient similarities of impressions between all approximated odors and their target ones were confirmed by a human panel.