Skip to Main content Skip to Navigation

Is it possible to discriminate odors with common words ?

Abstract : Several experiments have been performed in order to study the cognitive processes which are involved in odor recognition. The current report summarizes experimental protocol and analyzes collected data. The goal is to try to recognize odors from descriptors which are selected by subjects from a list. Different groups have to choose in several descriptor lists, some with profound descriptors and some with a few surface descriptors. Profound descriptors are supposed to involved more cognition than surface descriptors. Subjects also have to name the odors. Recorded data are first analyzed, and then learned by an incremental neural classifier. The problem is hard to be learned. It seems very difficult to discriminate the different odors from the sets of descriptors. A variant of the learning algorithm, less sensitive to difficult examples, is proposed. The pertinence of surface descriptors is discussed.
Document type :
Complete list of metadatas

Cited literature [4 references]  Display  Hide  Download
Contributor : Colette Orange <>
Submitted on : Wednesday, April 17, 2019 - 9:14:22 AM
Last modification on : Wednesday, May 27, 2020 - 12:32:03 AM


Files produced by the author(s)


  • HAL Id : hal-02102105, version 1



Regina Eberhardt, Helene Paugam-Moisy, Didier Puzenat, Jean-Pierre Royet. Is it possible to discriminate odors with common words ?. [Research Report] LIP RR-1995-31, Laboratoire de l'informatique du parallélisme. 1995, 2+16p. ⟨hal-02102105⟩



Record views


Files downloads