Signal Preprocessing And Identification Of Electronic Nose System

The methods are commonly used in the field of gas identification that sensory evaluation, analytical chemistry and gas sensor detection. The experimental results may lose objectivity using sensory evaluation, because it is affected by man’s subjective and olfactory system; the experimental process is more complex and the equipment is more expensive when the gas identification method is analytical chemistry; it is difficult to detection of gas using a single sensor in the complex environment which the gas is more than one. Then the electronic nose detection method is proposed, the basic components of electronic nose are gas sensor array. Not only the subjectivity and damage of using sensory evaluation, but also the complexity and expensiveness using analytical chemistry are overcame using electronic nose technology. In addition, the sensor array technology is used in electronic nose, which has some good merits as high sensitivity, strong real-time performance, simple operation, easy to carry, low cost, accurate and reliable. Therefore, it is important for gas detection to research of electronic nose technology.The main researches are hardware and software systems in the field of electronic nose. The manufacturing of sensor array is mainly research in hardware system, the software system mainly includes the preprocessing algorithm and pattern recognition algorithm. In the paper, the research focuses on preprocessing algorithm and pattern recognition algorithm. Firstly, some the theories of preprocessing and pattern recognition algorithms are introduced; the Parallel Factor Analysis algorithm (PARAFAC) is proposed as the recognition algorithm in electronic nose system according to the format of the data collected by the hardware equipment, and its theory are introduced. Secondly, the pattern recognition software system is built using the GUI of MATLAB, the functions of system include as following:original data import, data export; the preprocessing algorithm is built, that includes baseline processing, feature extraction, centering and scaling et al; and the pattern recognition algorithm is established; the results are showed by drawing and so on. Finally, the validity of the software system is analyzed by licorice data, and the experiments are done by the same preprocessing algorithm respectively combining with PARAFAC and PARFAR2, then the results compare with the results of principal component analysis algorithm. The results show that the parallel factor have some advantages in dealing with three dimensions data, and the improved parallel factor algorithm can effectively handle time shift problems, and make results more better. It achieves the research purpose of preprocessing analysis and recognition algorithm.