Automatic Dietary Monitoring

Contact

Dr. Oliver Amft

Project Start: August 2004

Objectives

Appropriate nutrition is a major aspect of a healthy lifestyle. Worldwide, a considerable number of life-years is lost due to the increasing prevalence of overweight, obesity and related chronic diseases. We believe that novel ubiquitous sensing solutions for continuous activity monitoring are vital cornerstones for diet behaviour coaching in risk groups and reduce diet-related diseases.
Current monitoring solutions require the user/patient to fill in questionnaires for all food consumed throughout the day. This is laborious and known to be very inaccurate. We develop and test solutions that do not require manual logging, hence Automatic Dietary Monitoring.

For this purpose we deploy sensors of different modalities that can be worn at the body or integrated into clothing. Pattern recognition methods are used to extract information and infer diet-related activities and consumed foods. Moreover we analyse sensor data for  information related to the eating microstructure. This can be used to broaden the understanding of malnutrition.

Achievements

We were able to demonstrate the feasibility of Automated Dietary Monitoring in a set of studies using on-body sensors and adapted pattern recognition methods. To capture the complex (partly unconscious) behavioural process we decomposed eating activities into different domains:

• Movements to manipulate and manoeuvre food (feeding gestures)
• Chewing, recorded from bone-conducted food breakdown sounds
• Swallowing  related to food intake.
  

An intake cycle model was defined and applied to segment feeding (intake) cycles, obtain insight on the eating process and determine its temporal structure. More...

Publications

• Gesture Spotting with Body-Worn Inertial Sensors to Detect User Activities.
  Holger Junker, Oliver Amft, Paul Lukowicz and Gerhard Tröster. Pattern Recognition, 2008, 41(6), 2010-2024.
  
• Recognition of Dietary Activity Events using On-body Sensors.
  Oliver Amft and Gerhard Tröster.  Artificial Intelligence in Medicine, 2008, 42(2), 121-136.
  
• Automatic Identification of Temporal Sequences in Chewing Sounds.
  Oliver Amft, Martin Kusserow, Gerhard Tröster. In BIBM 2007: Proceedings of the IEEE International Conference on Bioinformatics and Biomedicine, Fremont, CA, USA, 2007.
• Probabilistic parsing of dietary activity events.
  Oliver Amft, Gerhard Tröster. In BSN 2007: Proceedings of the International Workshop on Wearable and Implantable Body Sensor Networks, Aachen, Germany, 2007, Springer IFMBE Proceedings, vol. 13, 242-247. Pictures: 1 2
• Methods for Detection and Classification of Normal Swallowing from Muscle Activation and Sound.
  Oliver Amft, Gerhard Tröster. In PHC 2006: Proceedings of the First International Conference on Pervasive Computing Technologies for Healthcare, Innsbruck, Austria, 2006.
• Analysis of Chewing Sounds for Dietary Monitoring.
  Oliver Amft, Mathias Stäger, Paul Lukowicz, Gerhard Tröster. In UbiComp 2005: Proceedings of the 7th International Conference on Ubiquitous Computing, Tokyo, Japan, September 11-14, 2005, pages 56-72, Springer.
• Automated human dietary monitoring by fusing physiologic information.
  Oliver Amft, Gerhard Tröster. Poster at CIHP 2005: Symposium of the Center for Integrative Human Physiology (CIHP), Zurich, Switzerland, September 29-30, 2005.
• Detection of eating and drinking arm gestures using inertial body-worn sensors.
  Oliver Amft, Holger Junker, Gerhard Tröster. In ISWC 2005: Proceedings of the 9th IEEE International Symposium on Wearable Computers, Osaka, Japan, October 18-21, 2005.

International workshops

• e-Nutrition 2007: Workshop on Nutrition Monitoring and Public Health. e-nutrition 2007, Innsbruck, A, 2007 (co-located with Ubicomp 2007).

Media & Press coverage

• Medizinische Anwendungen im Wearable Computing. Austrian TV broadcast in Newton series (ORF1), September 3, 2006.
• Tragbarer Kontrolleur entlarvt Esssünder. Swiss Sunday newspaper, (SonntagsZeitung), February 19, 2006, page 77.
• Esssensoren: Elektroniker der ETH Zürich erfassen Essgewohnheiten mit neuartiger Sensortechnik. TV broadcast in MTW series (SF1), December 15, 2005.

Links

• Project insight page
• My homepage
• Workshop on e-Nutrition and Public Health (e-Nutrition 2007)
• Workshop on Health Coaching Systems 2008