Prediction of food texture changes using force data in simulations of repetitive chewing

Abstract

Food texture is a major factor contributing to the palatability of solid foods. The physical properties of food change due to crushing in the chewing process, hence the texture also changes. Food companies need to evaluate food texture in the chewing process. This study proposes a method to predict changes in texture during the chewing process. The developed method infers the dominant texture from multiple textures at a time point in the chewing process using a multinomial distribution within the framework of a state-space model. The model inputs are the texture parameters determined by Texture Profile Analysis from the repeatedly compressed measurement data of two successive times, and the output is the dominance rate determined by Temporal Dominance of Sensations. A dataset of measurement data and sensory evaluation data was compiled for using five foods and ten texture descriptors through measurement experiments and sensory evaluations. The effectiveness of the proposed method was verified by a cross-validation method. Root mean squared errors for the dominance rate of seven textures were less than 0.1. This study confirmed that the proposed method has the potential to predict food texture in the chewing process.