Research Of Implementation And Application Of Smart Pattern Recognition In Smartongue

Fast, real-time, on-line inspection of food quality and safety has been one of the main developmental goals in food quality control. Electronic tongue was developed in the mid-eighties of the twentieth century. It serves as a novel instrument to analyze and recognize the whole character of the liquor with the features of being fast, non-destructive and on-time which make it suitable for the research and application in the food quality control. Electronic tongue mainly includes three parts: (1) sensors array with low-selectivity, non-specificity and stable responding signal; (2) the equipment for actuating signal excitation and responding signal collection; (3) suitable pattern recognition or multivariate analysis method. Pattern recognition, as one of the three main parts in electronic tongue, plays a crucial role in the application and the research of electronic tongue. The complexity of the detected objects and the characteristic recognition feature and ability of each recognition algorithm make it true that no universal recognition method existing for all the cases. Therefore, the research on pattern recognition method for the Smartongue seems to be extremely important for its improvement.Smartongue, based on the pulse-voltammetry electrochemical method and the inertia nude metal electrodes, is developed by our laboratory as a novel electronic tongue which is attributed to the original concept of combinatorial pulse relaxation spectrum. In this work, Principle Component Analysis, Discriminant Function Analysis and Artificial Neural Network (ANN), the three representative methods of linear and non-linear pattern recognition, were used to evaluate the results from milk, tea beverage and Chinese liquor detected by Smartongue.The results are as follows:Firstly, the algorithm of Principle Component Analysis is realized by Matlab 7.1 software. PCA can do help Smartongue to discriminate the milk, tea beverage and Chinese liquor quickly.Secondly, Discriminant Function Analysis was introduced as a new kind of pattern recognition into Smartongue and Matlab 7.1 software was used to realize its algorithm. The result showed that Smartongue fulfilled the preliminary application of Discriminant Function Analysis in quality control of the milk, tea beverage and Chinese liquor.Thirdly, we explored the application of ANN in Smartongue and realized the algorithm of Competitive Neural Network and Probabilistic Neural Network. The recognition rate for milk, tea beverage, Chinese liquor of different flavor, different Chinese liquor of Luzhou-flavor, different Chinese liquor of Fen-flavor and different Chinese liquor of Rice-flavor is 91.7%, 89.6%, 79.2%, 78.6% and 97.2%, respectively by Competitive Neural Network, and 100%, 100%, 100%, 97.6% and 100% respectively by Probabilistic Neural Network. Probabilistic Neural Network is much more suitable to Smartongue pattern recognition than Competitive Neural Network.The implementation of data processing software set the stage for the following study.