Electronic nose and tongue combination for improved classification of Moroccan virgin olive oil profiles

Weakness of the recognition of odour and taste of similar patterns by electronic sensing systems has been recently investigated. This research was aimed at resolving this shortcoming by proposing an improved data fusion technique for a low-cost electronic nose combined with a simple electronic tongue system used to characterise five virgin olive oils (VOOs) from different geographical areas of Morocco. The principal component analysis (PCA) has shown a slight overlapping of the responses for the sensor array and indicated that no clear VOO discrimination can be drawn when we used the two instruments singly. Low-level of abstraction data fusion approach of the two systems has demonstrated the capability of discrimination that is superior to the two instruments taken separately. Furthermore, the selected features based on analysis of variance (ANOVA) led to an increase in the discrimination performance of VOOs. Afterwards, PCA and cluster analysis (CA) were carried out on the optimal selected variables and enabled an improved classification of the VOOs than using all the features. Therefore, the supervised method such as the support vector machines (SVMs) was applied to the new subset and confirmed that all VOOs were correctly identified according to their geographical origins. Results obtained with the improved data fusion approach outperformed the classification results of the electronic nose and tongue taken individually. (C) 2013 Elsevier Ltd. All rights reserved.