A feature extraction method based on wavelet packet analysis for discrimination of Chinese vinegars using a gas sensors array

A feature extraction method is proposed for discriminating three kinds of Chinese vinegars based on a gas sensor array composed of 13 Taguchi gas sensors (TGS). It employs three-scale wavelet packet analysis to decompose each signal of the sensor array into eight difference frequency bands, and the feature values can be obtained by computing the maximum of relative energy corresponding to each frequency band. Using the method, feature vectors of 13 dimensions were extracted from response signals of the array. At the same time, principal component analysis (PCA) and radial basis function neural network (RBFNN) were also employed to analyze these data so as to verify the validity of the method. The result of data processing indicated that both PCA and RBFNN could correctly discriminate the three kinds of vinegars. Therefore we think the feature extraction method is effective in respect of vinegars discrimination.