A Fast Method for Chinese Liquor Recognition

For the recognition of Chinese liquors, the existing methods are complicated and difficult to be applied to the electronic nose (e-nose)systems. To solve this problem, a novel liquor recognition method (coding method)is proposed in this paper. Ten gas sensors were used in our e-nose system. First, for each sensor response curve, four voltage features (called basic features)are extracted and combined into 36 new features (called combinatorial features)through the operations of addition and subtraction. And then the features of ten sensors are sorted from small to large. Finally, the (common)sensor orders of 40 features (4 basic plus 36 combinatorial features)are integrated as the (liquor)sample's code vector, and the testing sample class can be determined once the most similar liquor code vector (which means the maximum ratio of the identical code numbers between the testing sample and the liquors to the liquor code numbers)to that of the testing sample is found. To examine the performance of our proposed method, two different tests were carried out: the classification of five kinds of liquors and the true-fake liquor recognition. Classification results show that a good average accuracy of 97.3% is obtained using the proposed method. Especially, the accuracy of several liquors can reach 100%. The true-fake recognition results show a better average accuracy of 99.5%. Considering the classification (or recognition)accuracy and the simplicity of the proposed method, we have thus provided an useful method for the fast recognition of multiple Chinese liquors.