Differentiation of fresh seafood products and storage time using an electronic nose: features selection for data analysis
Electronic nose devices are becoming a practical tool for the identification of differences among food products badsed on their origin, intrinsic quality and process conditions. The difficulty in using such instruments is to obtain, from the sensors response patterns, suitable parameters to identify...
| Main Authors: | , , , |
|---|---|
| Other Authors: | , , , |
| Format: | Article in Journal/Newspaper |
| Language: | English |
| Published: |
2007
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/11585/58285 https://www.chiriottieditori.it/ojs/index.php/ijfs/issue/view/31/IJFS183-SLIM2006 |
| Summary: | Electronic nose devices are becoming a practical tool for the identification of differences among food products badsed on their origin, intrinsic quality and process conditions. The difficulty in using such instruments is to obtain, from the sensors response patterns, suitable parameters to identify the differences among samples. In this work the electronic nose analyses, trymethilamine determination and panel test were performed on samples of three species of farmed fish []ecentrachus labrax, Sparatus aurata and Salmo salar]tored for 8 days both under ice and abuse temperature of 6°C. Data from electronic nose were analysed choosing a group of features, i.e. average of the last 5 points, maximum point, average of the whole points and average of the signal resposnse characterised by the highest variation. A polynomial regression was performed on the different features in order to identify the contribution of the sensors to the sample response patterns. The data from the best sensors were then used to create an optimum feature (mixed feature). A principal component analysis (PCA)was performed on the data of all the features and the degree of separation of the tridimensional scatterplots used as a selective tool. Based on the results obtained, the features "average of the last 5 points" and "mixed feature" gave a good discrimination of the samples, "mixed feature" being the best. The scatterplot from "mixed feature" data allowed the separation of the fish samples based on species, freshness and storage time. The evolution overtime of the scores of the PCA, accounting for the highest variation of the best two features, was correlated both with TMA content and the overall evaluation of the product given by panelists. |
|---|