Premium
Integrating NIR Spectroscopy and Electronic Tongue Together with Chemometric Analysis for Accurate Classification of Cocoa Bean Varieties
Author(s) -
Teye Ernest,
Huang Xingyi,
Takrama Jemmy,
Haiyang Gu
Publication year - 2014
Publication title -
journal of food process engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.507
H-Index - 45
eISSN - 1745-4530
pISSN - 0145-8876
DOI - 10.1111/jfpe.12109
Subject(s) - electronic tongue , principal component analysis , pattern recognition (psychology) , cocoa bean , artificial intelligence , near infrared spectroscopy , normalization (sociology) , sensor fusion , mathematics , chemometrics , multivariate statistics , computer science , statistics , machine learning , food science , chemistry , biology , neuroscience , sociology , fermentation , taste , anthropology
This article examined the potential of sensor fusion of near‐infrared spectroscopy ( NIRS ) and electronic tongue ( ET ) together with multivariate analysis, for the accurate and rapid classification of five cocoa bean varieties. Optimum data extraction was done from each sensor and principal component analysis was used for data fusion by normalization. Support vector machine was used to develop the classification model. The model was optimized by cross‐validation and assessed by the numbers of principal components ( PCs ) and the classification rate. The single sensors ( NIRS and ET ) has a classification rate between 83 and 93%, while, data fusion ( ET‐NIRS ) had a classification rate of 100% at three PCs in both the training and prediction. Comparatively, the data fusion technique was superior to the single techniques. The findings could be exploited for reliable and rapid identification and discrimination of cocoa bean varieties. The study showed its novelty in the possibility of combining ET and NIRS data for accurate classification of cocoa bean varieties. Practical Application Cocoa processor will always demand high‐quality cocoa beans and breeders are in the business of releasing new varieties. The analytical methods used to assess and differentiate varieties are cumbersome, time consuming, and require an elaborate sample preparation and chemical usage. Also, single‐sensor technique does not also always provide accurate means to differentiate very similar varieties. This research investigated the feasibility of integrating near‐infrared spectroscopy ( NIRS ) and electronic tongue ( ET ) together with multivariate calibration analysis for rapid classification of cocoa bean varieties. This research has demonstrated that data fusion of NIRS and ET together with support vector machine could be used to differentiate cocoa bean varieties. These findings would be very useful to cocoa‐producing countries, processors, and quality assurance managers for overcoming mislabeling and adulteration. Breeders can also use this technique for rapid and easy differentiation of cocoa bean varieties to facilitate breeding process