Asian-style food intake pattern estimation based on convolutional neural network

Monitoring food intake behavior is a great help for people who need to manage or prevent obesity. Many researchers have proposed an automatic food intake monitoring technologies based on an image-based method and an inertial sensor-based method and so on. However, most previous works require additional equipment such as a camera on the table, ear clip type sensor with a microphone. Also, previous researchers have not developed on recognition methods considering various food intake cultures such as eastern and western. In this paper, we develop an algorithm that recognizes various food intake patterns, especially Asian cultures with accelerometer sensor without the need for additional sensors outside. We use three axis accelerometer sensors of a wearable device and apply convolutional neural network (CNN) to recognize various kinds of food intake behaviors such as eating with a spoon, picking food with chopsticks, and drinking water. The proposed model was verified by acquiring data from 8 subjects and showed 87.98% accuracy.