【文章內(nèi)容簡介】 analysis is a kind of subjects were divided into relatively homogeneous groups of statistical analysis techniques. From a statistical point of view, the cluster analysis is a way to simplify the data through data modeling. CA is classified based on the relationship between the number of the individual or variable the objectivity strong, but various clustering methods can only be achieved under certain conditions the local optimum。 whether the final result of the clustering to set up, take identification of experts.Discriminant factor analysis (DFA)The judgment factor analysis is a statistical method to determine the individual category. Observations according to the known classes of two or more samples to determine one or more linear discriminant function discriminant index, then another body to determine which category the discriminant function based on discrimination index. DFA according to the results of the standard sample used for electronic noses to identify blind. By rebination of sensor data to optimize the differentiated, or through the sensor optimized selection, . removal of the sensor, no contribution or a small contribution to improve the recognition capability, its purpose is to bring the group distance between the difference within a maximum while ensuring group minimum.Artificial Neural Network (ANN)The artificial neural network is a by imitate human or animal neural network behavioral characteristics, mathematical model of distributed parallel information processing. This network relies on a system of plex procedures, by adjusting the the internal large number of interconnected relationships between nodes, so as to achieve the purpose of processing information. ANN provided in advance a number of mutually corresponding input data output data, analysis, grasp the potential between the law, and ultimately with a new input data According to these laws, to derive output, this The learning process of analysis is called training. The aforementioned methods, ANN is usually considered to be a promising approach, and its features and benefits mainly in three aspects: with selflearning and adaptive function。 associative memory function。 with highspeed to find the optimal solution capacity. In addition, it is able to solve nonlinear problems better than the traditional statistical methods in dealing with noise and drift. Currently, many artificial neural network is used for processing the signal of the sensor array, such as BP neural network, radial basis neural networks, fuzzy neural networks, selforganizing network.圖2Research progress:In 1964, Wilkens and Hatman use of gas on the electrode, the oxidationreduction reaction olfactory process electronic analog, which is the earliest reports on the electronic nose. 1965, Buck et al the use of changes in the conductance of a metal and a semiconductor gas measurement, Dravieks then using the change in contact potential measurements of the gas. However, as the gas classification with the concept of intelligent chemical sensor arrays until 1982 by the University of Warwick, UK Persuad et al, their electronic nose system consists of two parts of the gas sensor array and pattern recognition system. The sensor array part of the mountain three semiconductor gas sensor. This simple system can distinguish between the brain, according to the tree rose oil, clove oil, volatile chemical substances for dental odor. In the next five years, the electronic nose research and did not cause extensive attention in the international academic munity. 1987, held at the University of Warwick, UK 8th Annual Meeting of the European chemical sensing research organization is the turning point of the study of the electronic nose. At this meeting, Gardner led by the University of Warwick in gas sensing sensing research team published a paper sensor in gas measurement parties and applications, focusing on the concept of pattern recognition, caused academia extensive interest. In 1989, the North Atlantic Treaty Organization held a special chemical sensor information processing symposium, dedicated to artificial olfactory system design of these two topics. In August 1991, the North Atlantic Treaty Organization was held in Iceland thematic sessions of the first electronic nose. The electronic nose Since then rapid development. 1994, Gardne: a review article on the electronic nose, formally proposed the concept of electronic nose, marking the electronic nose technology into the mature stage of development. Since 1994, after more than ten years of development, the electronic nose research has made rapid progress. For the study of the electronic nose mainly in the side of the sensor and electronic nose hardware design, pattern recognition theory, the electronic nose in food, agriculture, pharmaceutical, biofield applications, electronic nose, and biological systems while. The hardware design of the sensor and the electronic nose and electronic nose applications in the field of food and agriculture is a hot spot in the study of the electronic nose. Crewe town in Cheshire, England pany successfully developed an electronic nose, the tests showed that the patient skin wound bacteria can sniff out the erosion remind doctors to take appropriate measures in a timely manner. His study of the electronic nose is a matrix posed of 32 different organic polymer sensor, very sensitive to the smell of various volatile pound, the pound is different, the reaction is different. Typically, bacterial growth will be issued chemical odor the electronic nose contact with the odor, each sensor resistance will change. Format Since each sensor corresponding to a different kind of chemical substances, posed of 32 kinds of the same change in resistance would respectively represent the fingerprint of the different odors. The tests showed that the electronic nose can be found if there is the presence of bacteria only need a few hours. In the past, using the met