Application of ANN with extracted parameters from an electronic nose in cigarette brand identification

Abstract

The Chinese cigarette industry is losing millions of dollars per year due to counterfeit cigarettes. Detecting illegal cigarettes in the field is difficult, but may be possible using portable electronic noses (E-noses). However, the differences between odours from counterfeit and genuine cigarettes are small and detection may prove difficult. In this study we propose a practical approach to increase the performance of E-noses in cigarette brand identification. A portable E-nose was employed to collect and classify aroma signals from different brands of cigarettes. Artificial neural networks (ANNs) were employed and trained with raw data and extracted features from the data collected by the E-nose to identify the cigarettes. This preliminary investigation succeeded in identifying four different types of cigarettes in the laboratory. The identification results obtained from the neural network trained with extracted parameters were better than the ones obtained directly from the E-nose.