Prediction of Eating Activity using Smartwatch
Haruka Kamachi, Tahera Hossain, Fuyuka Tokuyama, Anna Yokokubo, Guillaume F. Lopez · 2021
Eating activity understanding has been extensively studied for its importance in our lives. Due to the lack of proper dietary management, obesity has been increasing worldwide, which causes many diseases. In this regard, adequate diet management is crucial for having a healthier life. This study aims to develop a system that allows users to manage their diet easily. This research proposed a method that estimates eating activities. The main objective of this research is to use an eating manner (way of eating) to identify more accurately the eating activity by arm’s acceleration using a smartwatch. The acceleration data were collected using a smartwatch during different mealtime. A total of 12 features are extracted from this data. We have examined different classifiers’ performance. Among them, the Support Vector Machine (cubic SVM) classifier performed best among all. An evaluation of the system resulted in eating activities detected with the maximum accuracy of 76.4% and an average accuracy of 71.9% using 10-folds cross-validation.