An Efficient Method for Odor Retrieval

Recently, researchers have been increasingly interested in odor database retrieval using the sense of smell. However, it has two difficulties; different from sight- or hearing-sense database retrieval. One is that odor scientists have not yet been able to find a base component of odor, such as RGB or frequency. Therefore, smell-sense database retrieval cannot be conducted using a physical quantity. The other is that relevance tests of each retrieval result require a larger load. Conventional approaches have represented an odor by either a noun or impression word. It is more feasible if a user can efficiently obtain the relevant retrieval results by employing both nouns and impression words such as 'odor like slightly sweet coffee'. In this paper, we propose such an efficient method.