Near infrared spectroscopy (NIRS), combined with multivariate calibration methods, has been used to quantify the robusta variety content of roasted coffee samples, as a means for controlling and avoiding coffee adulteration, which is a very important issue taking into account the great variability of the final sale price depending on coffee varietal origin. In pursuit of this aim, PLS regression and a wavelet-based pre-processing method that we have recently developed called OWAVEC were applied, in order to simultaneously operate two crucial pre-processing steps in multivariate calibration: signal correction and data compression. Several pre-processing methods (mean centering, first derivative and two orthogonal signal correction methods, OSC and DOSC) were additionally applied in order to find calibration models with as best a predictive ability as possible and to evaluate the performance of the OWAVEC method, comparing the respective quality of the different regression models constructed. The calibration model developed after pre-processing derivative spectra by OWAVEC provided high quality results (0.79% RMSEP), the percentage of robusta variety being predicted with a reliability notably better than that associated with the models constructed from raw spectra and also from data corrected by other orthogonal signal correction methods, and showing a higher model simplicity.