Advancements towards the development artificial intelligence for sensory analysis: Integrating pattern recognition and signal processing in ATR-FTIR analysis of spirits

Abstract

Spirits have been gaining acceptance by international consumers due to its distinct organoleptic profile, which provide each beverage with a characteristic sensation of the product (taste, smell, and color). However, it is important to establish differences by an analytical technique supported by artificial intelligence to demonstrate with mathematical solutions rather of being subjective such as human senses. To reduce the uncertainty associated to the sensory analysis, the present study shows as strategy the recognition of patterns of ATR-FTIR spectra of a variety of distilled beverages: Tequila 100% agave (n = 20), Mezcal (n = 20), Raicilla (n = 20), Aguardiente (n = 20), and Vodka (n = 20). All ATR-FTIR spectra present seven characteristic bands associated with methanol, ethanol, isoamyl alcohol, butanol, ethyl acetate, acetaldehyde, and furfural in beverages. Based on the analysis of the peak signals obtained from the Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR) spectra, the implementation of an Artificial Neural Network was proposed with the objective to classify each alcoholic beverage in its corresponding groups. Results showed that the artificial intelligence algorithm proposed permits to correct classify 100% of the samples being a great alternative strategy being innovative, fast, and of low-cost strengthening food quality analyses by machine learning decision-making.