Intake gestures and movement
The characteristic movement patterns during intake were first discussed in our 2005 paper on detection of eating and drinking gestures. The approach in most follow-up work has been to measure movements using inertial sensors attached to the wrist. Our work focused on developing spotting methods to identify different intake motion in continuous data streams from wearable sensors. The pattern spotting methods are based on hidden Markov models (HMM) and feature search strategies. The methods are generically applicable to different event detection and gesture spotting tasks in data streams. We further developed segmentation methods to leverage the intake-specific arm rotations in roll and pitch axes. The work on segmentation and spotting was published first in 2005 and incorporated in more elaborated form later. Results showed that different intake movements can be discriminated, including drinking, eating with fork and knife, with spoon, and using hands only. Subsequently, we investigated self-adaptation and sparse coding techniques to personalise and maximise intake movement spotting performance.
Our investigations continued to explore magnetic proximity measurement methods between hand and chest/head to support gesture segmentation and identification of different drinking movements related to the fluid container used.
Publications
[publication-list tags=”intake gestures” separator=0]
Contact
Prof. Dr. Oliver Amft
- Job title: Director
- Phone number: +49 9131 85-23601