Application of biologically modeled chaotic neural network to pattern recognition in artificial olfaction.

This paper presents a novel neural network called KIII model for pattern recognition in artificial olfaction, whose topological structure and parameters are based on anatomical and electrophysiology experiments in mammalian olfactory system. Six data sets of three volatile organic compounds in different conditions, each with a wide range of concentrations, are obtained by a signal acquisition system with tin oxide gas sensor array. They are input into KIII model for training and test. Experimental results show that the system had a good classification performance in a wide concentration range while only a few training samples needed.