The Identification Of Wine Based On PCA And Ann

Electronic nose (EN) is a equipment with ability of identification of simple or complex odors which mentioned gas sensor, multi-information data fusion and optimization, computer technology and applied mathematics and other fields. Because of its low cost and accurate identification rate, the researches of it attract more attention and it develops quickly.In this thesis, the principle and the structure of the EN aiming at the identification of wines, the characteristic of gas sensors, the design of signal collected circuit and the selection of pattern identification are analyzed.The EN software system is written by Visual C++ language. Through the Windows API functions to control serial port communication of computer and gas sensor array output circuit, a friendly interface is provided. It realizes modification of parameters of serial port, sampling interval and whole sampling time, and stores sampling data to a text file.The identification of three kinds of wines (spirit, beer,Tokay) and two brands of spirit is performed. At first sampling data is preprocessed and a better algorithm is chosen. Improved BP algorithm and nearest neighbor–K means clustering RBF algorithm is put forward. And the test result proves these two algorithms can identify the sampling data well.In addition, principal component analysis (PCA) method wipes off redundant sensor from the sensor array. Principle components and brief sensor signals are tested by above two algorithms, and they can be identified well, too.