A comparison between an electronic nose and human olfaction in a selected case study

An electronic nose is now becoming available as a commercial product. Nevertheless its performances are not fully understood and interpreted. Also the differences between electronic noses and the human olfaction have not yet been sufficiently studied. This is an important issue in many industrial sectors, such as food analysis. In this paper a comparison between the performances of an electronic nose and a panel of human tasters is presented in a selected case (tomato paste). An extensive set of tools for data analysis was available. A number of chemometrics based methods (principal component analysis and cluster analysis) and neural networks (feedforward backpropagation trained networks, self organizing maps, adaptive resonance theory based networks) have been utilized to analyze electronic nose data in order to extract the relevant information. The electronic nose and the human panel show strong similarities but the former displays a more concise classification capability for the data.