Recipe-Recommendation System Based on Physical-Characteristics Recognition Using a Depth Image Sensor
With the development of information technology, it has become easy to construct a system to recommend favorite
products to customers. Owing to the increase in lifestyle-related diseases such as high blood pressure or diabetes, a
nutriment-based recipe recommendation system is in great demand for national health. Based on Dietary Reference Intakes for
Japanese, a research to estimate nutriments per meal from physical characteristics and recommend a recipe has been performed.
A “recipe-recommendation system” that considers nutrients on the basis of a user’s physical characteristics acquired by depth
image sensor was designed, developed and evaluated. The system extracts the physical characteristics namely “age”, “gender”,
“height”, “weight”, and “body mass index (BMI)” in a short time and estimates the required amounts of nutrients per meal. It
then finds a combination of dishes that suit the user’s taste from a cooking database and recommends a menu of dishes to the
user. From experiments on recognition accuracy of the system, average recognition accuracy of user physical characteristics
was 89%. Moreover, the recipe recommendation based on physical characteristics acquired by depth image sensor was
confirmed to be sufficient.
Index Terms— Recipe-recommendation system, human-body recognition, depth image sensor, joint position, dietary
reference intakes, physical characteristics.