4 and 100%, for the model based on normalized spectra, and 94.6 and 95%, for the model based on first derivatives. Such results confirm that DRIFTS provides satisfactory discrimination between defective and non-defective roasted coffees, demonstrating its potential for detection of defective beans in mixtures with non-defective ones after roasting. Regarding the application of such methodology for routine analyses

of roasted coffee quality, further studies are still necessary, involving a trained panel of coffee tasters, to establish the minimum amount, if any, in which defective beans can be introduced to a non-defective coffee batch and changes in the beverage quality would still not be perceived in relation to one without defective beans. With the minimum amounts effectively established, mixtures of defective and non-defective roasted beans can be suitably prepared and duly tested for the find more discrimination capability of the developed models. The feasibility C59 wnt of employing DRIFTS as a methodology for discrimination between defective and

non-defective roasted coffees was evaluated. The obtained spectra were similar, with small differences in absorbance intensity between non-defective and defective coffees. PCA results based on DR spectra and first derivatives indicated separation of the samples into four major groups: non-defective, black, dark sour and light sour, with immature beans scattered among the sour samples. LDA classification models, based on absorbance readings and derivatives at eight wavenumbers (2924, 2852, 1743, 1541, 1377, 1076, 910 and 816 cm−1), provided separation of the samples into five groups: non-defective, black, dark sour, light sour and immature beans. Average recognition

and prediction abilities ranged from 79 to 96% and from 80 to 100%, respectively. Discrimination functions for generic classes of defective and non-defective coffee samples were also developed. For these generic models, recognition and prediction abilities ranged from 95 to 97% and from 95 to 100%, respectively. The results obtained in the present study confirm that DRIFTS provides satisfactory levels of discrimination between defective and non-defective coffee beans after roasting. The authors acknowledge financial support from the following Brazilian Government Agencies: CNPq from and FAPEMIG. “
“Events Date and Venue Details from COFE 2012 – 11th Conference of Food Engineering 2–4 April 2012 Leesburg, Virginia USA Email:[email protected] Food Colloids 2012 15–18 April 2012 Copenhagen, Denmark E-mail: Richard Ipsen: [email protected] NEFood: 1st North European Congress on Food 22–24 April 2012 St. Petersburg, Russia Internet:http://nefood.info 8th International Conference on Diet and Activity Methods 8–10 May 2012 Rome, Italy Internet:http://www.icdam.org 11th International Hydrocolloids Conference 14–17 May 2012 Purdue University, USA Internet:http://www.international-hydrocolloids-conference.