Abstract

Journal of Near Infrared Spectroscopy
Volume 12 Issue 3, Pages 183–188 (2004)
doi: 10.1255/jnirs.425

A hierarchical discriminant analysis for species identification in raw meat by visible and near infrared spectroscopy

Thorsteinn Arnalds,^a† John McElhinney,^b Tom Fearn^a and Gerard Downey^b
aDepartment of Statistical Science, University College London, Gower Street, London WC1E 6BT, UK^bTeagasc, The National Food Centre, Ashtown, Dublin 15, Republic of Ireland
^†Current address: Faculty of Engineering, Technical University of Iceland, Höfðabakka 9, 110 Reykjavík, Iceland

The problem tackled is that of discriminating between chicken, turkey, pork, beef and lamb using visible and near infrared spectra collected from homogenised meat samples. This five-group classification task was treated as a hierarchical sequence of binary splits: white versus red meat, then either poultry versus pork or beef versus lamb, and finally, in the case of poultry, chicken versus turkey. Most of the splits were achieved by linear discriminant analysis applied to principal component scores. The chicken versus turkey split worked better with soft independent modelling of class analogy (SIMCA). One of the attractions of the hierarchical approach is that different methods can be used for different splits. The results, with only two classification errors on 115 validation samples, were better than those achieved in previous analyses of the same data.

Keywords: near infrared spectroscopy, authenticity, meat, chemometrics, discriminant analysis, principal component analysis, canonical variate, SIMCA

---

Full-text article (426 kB) (subscribers only)

Buy article on-line for £20 (get immediate access)

Alerting Service

 RSS Feed

Permalink: http://dx.doi.org/10.1255/jnirs.425
QR Code (what is this?):

---