Turmeric powder is valued both for its medicinal properties and for its popular culinary use, such as being a component of curry powder. Due to its high demand in international trade, turmeric powder has been subject to economically driven and hazardous chemical adulteration. The objective of this study was to classify and quantify two major adulterants (Sudan Red and Metanil yellow) in turmeric powder using near-infrared (NIR) spectroscopy associated with chemometric models. The absorbance spectra of samples were obtained in frequency regions of 10000–4200 cm-1. A Soft Independent Modeling of Class Analogy (SIMCA) model was established to determine the authenticity of the turmeric powder. A partial least squares (PLS) regression was used to predict the adulteration level. The different combinations of preprocessing methods such as standard normal variate transformation (SNV), median filter, Savitzky-Golay, wavelet, multiplicative scatter correction (MSC), first derivative (D1) and second derivative (D2) were applied to preprocess the spectral data, in which SNV combined with median filter and D1 performed best. The sample correct classification rate in SIMCA model were found 96.7 and 93.4 in both calibration and validation set respectively. The PLS regression models were evaluated by the coefficient of determination (r2) and root mean square error (RMSE). All PLS regression models established with the combination of different spectral pretreatment except of MSC, Savitzky-Golay, D1 and MSC, Wavelet, D1 showed low values of RMSE and ideal coefficient of determination for both calibration and prediction sets while using relatively few factors. Also, the PLS regression results showed that first derivative led to better calibration models when compared with the second derivative. The obtained statistical parameters for Sudan Red were r2=0.90, RMSE=0.059 in test set and r2=0.91 and RMSE =0.058 for metanil yellow in test set.

Keywords