【正文】 the different regions, followed by the use of principal ponent analysis, to further reduce the number of dimensions. Development RBF neural networkbased classification model for feature extraction. electronic nose to detect the Pink Lady brand apple shelf life, results with routine quality testing (hardness, starch index, acidity, etc.) related contrast, showed a good correlation Supriyadi electronic nose of different maturity snakeskin fruit detection, results showed that the electronic nose can be a good distinction。 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。 ground state phase sampling stage, holding phase and the recovery phase. Stage in the base state, the threeway solenoid valve is switched to the air passage, and a voltage signal of the sensor array to a horizontal baseline. To be a threeway solenoid valve is switched to the sample gas channel into the sampling stage, the sensor array response, a few seconds after the voltage signal value rose to a peak and then stabilized, to be close to equilibrium, the system enters the holding phase, threeway solenoid valve remains constant sample gas channel. The last to enter the recovery phase, threeway solenoid valve is switched back to the air channel, rapid decline in the response curve of the sensor array, until it is restored to the baseline position. The time value of the four stages of the testing process before the start of the test, you must first set on a PDA: the ground state time, sampling time, hold time and recovery time. Then Start mand, the system began to test issued by the PDA. Collection terminal after receiving the order, the internal microcontroller in accordance with the instructions, timer to control the time of the four stages. First, the microcontroller will control the heating voltage of DAC conditioning module output is used to heat the sensor array, and then control the signal voltage of the ADC module acquisition sensor array and choose to get the data saved to the memory module, or directly to the receiver by the wireless module feedback PDA to do the analysis and processing. After the data acquisition is pleted, PDA sends end mand, the system stops working. During the next round of testing is required before the PDA issue a reset mand to make the system cleared before a new round of tests. In addition, in the process of testing, you can issue mands through the PDA miniature air pump and threeway electronic valve control switch. After the test is stopped, the data of the sensor array can be read through the open instruction to obtain before and response curves.Statistical methods:Principal ponent analysis (PCA)Principal ponent analysis refers to several variables by linear transformation to elect a fewer number of important variables of multivariate statistical analysis method, also known as principal ponent analysis。 2) Determination of the sample and data analysis: sampling operation unit the initialization of the sensor array is exposed to the odor body, when the contact with the surface of the active material of the volatile pounds (VOC) and the sensor (ultrasonic sensor), to produce the transient response. Such a response is recorded and transmitted to the signal processing unit for analysis, and stored in the database of a large number of VOC pattern parison, identification, and to determine the type of odor。 and to avoid human error, good repeatability。 Similarly, the 2nd gas generating high response of the sensor is not sensitive on the 1st gas and, ultimately, the entire sensor array response pattern is different for different gases, it is this difference, to enable the system to identify a gas according to the response of the sensor pattern. Electronic nose can be summarized as follows: the sensor array the signal preprocessing neural networks and a variety of algorithms puter identification (gas qualitative and quantitative analysis). Functionally speaking, the gas sensor array is equivalent to the biological olfactory system in a large number of olfactory receptor cells, neural network and the puter to recognize the biological equivalent of the brain, the rest is the equivalent of the olfactory nerve signal transduction system. For example, the measurement of the metal oxide semiconductor (MOS) gas sensor in response to th