A bio-inspired data processing method for classification of Chinese liquors using electronic nose

We proposed a bio-inspired neural network to perform data processing in an electronic nose (e-nose) system. The structure of the proposed neural network is similar to mammals' olfactory system which contains olfactory sensing neurons, mitral cells and granule cells. This neural network largely simplifies traditional data processing steps in e-noses. It uses the original data collected from gas sensors without data preprocessing and feature reduction. We innovatively use recurrence quantification analysis to perform feature selection. Recurrence rate and determinism are calculated according to the network's output curves and these values are used as features extracted from the proposed olfactory neural network. Linear discrimination analysis and support vector machine are used as classification methods. The performance of the proposed neural network was tested by “leave-one-out” cross-validation. The classification rates were 93.75% and 96.25% using these two methods, respectively, which were higher than traditional data processing methods (85% and 91.25%).