Optimization of electronic nose measurements and discrimination of apple juices in combination with supervised pattern recognition

A common theme of food authentication studies is the requirement to which the raw apple juice samples can be compared to establish its authenticity. In order to discriminate eight varieties of apple juices using electronic nose (EN), we tried to extract the most relevant information from EN response signals. Experiment parameters were optimized to ensure the response curve fully characterizing sample information. The optimal conditions were 10 mL of volume and 90 min equilibration. Sensor optimization was conducted to eliminate redundant information. Sensors W1C, W5C, W3S, W2S, W5S and W1W were chosen for pattern recognition. This process improved the PCA-based pattern performance. Best discrimination performance was obtained utilizing response signals of stationary phase according to multivariate analysis. Linear discriminant analysis (LDA) and Support Vector Machine (SVM) were carried out to develop discrimination models. Both LDA and SVM achieved satisfactory variety-based classification performance, with both 100% accuracy classification rates in terms of recognition ability and prediction ability. The perfect performance indicated that EN can be successfully applied in apple juices discrimination. And the results obtained will play a positive role in successful commercialization application of EN to apple juice industry.