Discrimination of civet coffee using near infrared spectroscopy and artificial neural network

摘要

Civet coffee is regarded as the most expensive coffee in the world. Owing to its high price the conventional method of harvesting civet coffee in the wild has been replaced by a farming method, wherein the wild civet cats are being captured, caged, and then force-fed with ripe, hand-picked, coffee cherries. This is contrary to the production of wild civet coffee from the ripest and sweetest coffee cherries, carefully picked by the civet cat. As the wild civet coffee is quite hard to obtain, most of the civet coffee commercially sold as authentic civet coffee is actually from a caged civet cat. Traders and consumers have no way of differentiating the civet coffee from other types of coffee. This study aimed to differentiate caged civet coffee beans from ordinary green coffee beans. The technique used was the near infrared spectroscopy (NIRS) as it is non- destructive and can generate quick results. A total of 218 samples were scanned, which generated absorbance in the entire 780 wavelength capacity of the Cavite State University Indium, Gallium and Arsenic (CvSU InGaAs)-based NIR instrument, ranging from 904 to 1684 nanometres (nm). A total of 170,040 spectra were generated. The average spectral absorbance’s having major differences between the two groups, were chosen, which are 907nm, 1088 nm, 1540 nm and 1650 nm, respectively. Out of 218 samples , 130 samples were used as training data, 40 samples as testing data, and the remaining 48 samples for validation purposes. The training data were subjected to the 4 layers, 15 neurons feed forward back propagation artificial neural network (FFBPANN) for training. Classification scores of 95% to 100% were achieved. Using the combined NIRS and FFBPANN, the civet coffee can be successfully discriminated from coffee beans not eaten by a civet.