Monitoring of Human Eating Behaviour in a Restaurant

Abstract of the Project

Analysis of a person’s food selection and food intake activities can help in monitoring the person’s diet and nutrition. The goal is to develop an unobtrusive context aware system for in-time personalised food selection support for regular eating environments. Multimodal technology would be used for automatic human behaviour measurement, reasoning and feedback for a variety of application domains. This project contributes to the ‘Diet & Nutrition’ application field among the Healthcare, Wellness and Assisted living (HWA) applications.

The research focuses on measuring human eating behaviour in a natural environment like the Restaurant of the Future (RoF) at Wageningen University which is  a ‘living laboratory’ for unobtrusive studies of food choice and eating behaviour. The research focuses on finding an ideal combination of video camera and other environmental sensor observations. As a start, a vision-based method is developed to recognize eating behaviour of a person using a new dataset recorded to provide the training and testing data. The method uses dense trajectories for activity recognition and is robust to fast irregular motions. It is suited for uncontrolled realistic environments.

Research Achievements to date

• iEatSet dataset recording in progress
• Vision-based activity recognition method for human eating behaviour being developed
• Systematic literature review

References

[1]  H. Wang, A. Klaser, C. Schmid, and C.-L. Liu, “Action Recognition by Dense Trajectories,” In: CVPR, 2011