This paper proposes an algorithm for recognizing known and unknown odors. The corresponding system developed by Widyanto et al. is called Euclidean fuzzy similarity-based self-organized network inspired by immune algorithm (EF-SONIA). This approach is an improvement over other recent systems, by using Euclidean fuzzy similarity.

The first section of the paper introduces odor discrimination systems and the motivation behind them. It also explains the sensor mechanisms involved. The second section describes F-SONIA in detail. The limitation of F-SONIA--that it cannot recognize unknown odors--is described mathematically. The third section shows how this limitation is overcome by EF-SONIA, by showing how Euclidean fuzzy similarity can expose unknown odors. It proposes two methods to calculate the involved fuzziness: one is to take the average of two fuzziness widths, and the other is to do the calculations as if the problem model is an ellipse. The fourth section describes the experimental results. Data sets for three known perfume odors are used as input. The proposed methods are found to be more accurate than F-SONIA in recognizing known odors. They are also found to be better than conventional methods in recognizing unknown odors. The fifth section concludes that the elliptical approach is the most accurate, but more time consuming.

The paper gives a good idea of the mechanical technologies involved in an odor-recognition system. The major contribution is explained using mathematical formulas. Performance is presented using tables.