Food texture analysis based on machine learning

Abstract

Food texture is one of the most important factors in deliciousness of the food, dominating more than 60%. Sensory analysis has been used mainly for evaluation of the food texture. Toward more stable evaluation, instrumental analysis, based on physical and chemical properties of food, has been applied. Characteristic values, such as peak value of the force in food compression test, reflects some aspects of food texture. However, it cannot evaluate the food texture in many food, due to the complexity of food consisting of muti-component, and inhomogeneous structure. In this study, all the raw force values obtained in food compression measurements were used to analyze food texture. Using machine learning, some snacks such as potato chips, which cannot be analyzed by conventional methods based on multivariate analysis, were successfully classified based on food texture, and the classification accuracy ranges from 0.7~0.9.