Rapid and nondestructive detection of multiple adulterants in kudzu starch by near infrared (NIR) spectroscopy and chemometrics

Abstract Traditional analysis of food adulterations is usually focused on detection of specific adulterants that are known or suspected to exist. However, because there are various ways of adulterations and new adulterants are constantly emerging, the usefulness of the traditional analysis methods is limited. This paper investigated the feasibility of using near infrared (NIR) spectroscopy and chemometrics to detect a range of adulterants in a Chinese functional food, kudzu starch (KS). Pure KS objects ( n  = 125) were collected from five main producing areas. Five potential adulterants were investigated, including an illegal but commonly used whitening agent, talcum powder, and four cheaper plant starches, namely, sweet potato, potato, maize and cassava starches. The doping levels were 2%, 4%, 6%, 10% and 20% (w/w) for plant starches and 0.5%, 1%, 2%, 5% and 10% (w/w) for talcum powder. A recently proposed one-class classifier, partial least squares class model (PLSCM) was used to estimate the distribution of authentic KS samples based on NIR spectra. The results indicate that taking second-order derivative (D2) spectra can effectively improve the PLSCM model specificity for detecting adulterants. D2-PLSCM can detect 0.5% or more of talcum powder and 4% or more of the four cheaper starches.