Abstract:Introducing some principles of biological pattern recognition into artificial olfaction will produce more biomimetic instrument, and they are also considered as promising sensor array information processing methods. This paper presents an electronic nose based on an olfactory neural network to detect and discriminate green tea with different brands. Data of five brands of green tea from different areas were acquired through a homemade electronic nose apparatus containing eight MOS gas sensors. Feature vectors extracted from steady-state phase of array signals, were plotted on radar charts as fingerprint, to inspect the effectiveness of sensor array arrangement and feature selection method. The olfactory neural network trained by a biologically plausibility learning rule was used to classify those samples. The relationship between training cycles and recognition rate was investigated. The results showed that after 4~7 training cycles the network could correctly discriminate those green teas with average recognition rate of above 97%.