Food texture analysis based on deep learning of mastication 3D behavior collected by smartphone 3D-scanner

Abstract

Chewing is the one of the most basic and important action for humans to live. Since a series of different muscles exist and work together around mouth during mastication, it is difficult to measure precisely their movements. In this study, we tried to collect big data of the mastication against a series of different foods by measuring 3D shape of the face as a function of time during mastication based on 3D scanner in the smartphone. After deep learning of this mastication big-data based on the label of food characteristics, such as type of the food, hardness of the food, a series of foods were able to analyze based on this method. For example, some physical characteristics, type of foods, and the amount of the food were precisely predicted based on the 3D facial surface motion. Even for the foods having similar food texture, this learned model based on mastication data can distinguish and detect the difference the foods having similar food texture. This measurement and analysis methods could be applied not only for the food texture analysis but also for the dental field such as food education.