This paper addresses research challenges associated with development of a wearable sensor system for detecting the food consumption activities of a subject. The objective is to automatically detect the occurrence of food consumption whenever it occurs, in order to use this activity detection to record a representative camera image of the food and count the number of bites of food consumed. The wearable system consists of two elastic bands - one each on the upper arm and wrist - instrumented with wireless inertial and magnetic sensors. Two major technical challenges include i) singularity issues with Euler angle estimation due to arm rotations that can exceed 90 degrees, and ii) the need to differentiate between eating and non-eating activities involving close hand-mouth proximity. The singularity challenge is addressed by using a direction cosine matrix estimation technique that utilizes a linear Kalman Filter. The differentiation between eating and non-eating activities is done using a support-vector-machine (SVM) based machine learning algorithm. Experimental results using wearable prototype bands show that both the DCM estimation and machine learning components work reliably and have the potential to be useful for home-based automated food intake detection.
|Original language||English (US)|
|Title of host publication||2019 American Control Conference, ACC 2019|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Number of pages||6|
|State||Published - Jul 2019|
|Event||2019 American Control Conference, ACC 2019 - Philadelphia, United States|
Duration: Jul 10 2019 → Jul 12 2019
|Name||Proceedings of the American Control Conference|
|Conference||2019 American Control Conference, ACC 2019|
|Period||7/10/19 → 7/12/19|
Bibliographical noteFunding Information:
*Research partially supported by the Digital Technology Center, University of Minnesota.
© 2019 American Automatic Control Council.