Effects of sensory combination on crispness and prediction of sensory evaluation value by Gaussian process regression

Abstract

Crispness contributes to the pleasantness and enjoyment of eating foods and is popular with people of wide ages in many countries. Hence, a quantitative evaluation method for crispness is required for food companies developing new food products. In this study, the effects of different sensory combinations on crispness were investigated through sensory evaluation, and a Gaussian process regression model was used to predict the evaluation values of crispness. First, four crispness descriptors in Japanese were selected, and sensory evaluations were conducted with ten participants using commercially available snack foods under three different sensory combinations of force, vibration, and sound to confirm the effects of the three senses. An instrumental system also measured force, vibration, and sound for snack foods under the same conditions. The Gaussian process regression model determined the relationship between the sensory and measurement data and predicted the sensory evaluation values from the measurement data. Cross-validation verified that the Gaussian process regression model accurately predicted the food texture evaluation values from the measurement data even in conditions with different sensory components.