Date of Award

August 2020

Document Type


Degree Name

Doctor of Philosophy (PhD)


Computer Engineering

Committee Member

Adam Hoover

Committee Member

Brian Dean

Committee Member

Yingjie Lao

Committee Member

Yongqiang Wang


Eating is one of the most basic activities observed in sentient animals, a behavior so natural that humans often eating without giving the activity a second thought. Unfortunately, this often leads to consuming more calories than expended, which can cause weight gain - a leading cause of diseases and death. This proposal describes research in methods to automatically detect periods of eating by tracking wrist motion so that calorie consumption can be tracked. We first briefly discuss how obesity is caused due to an imbalance in calorie intake and expenditure. Calorie consumption and expenditure can be tracked manually using tools like paper diaries, however it is well known that human bias can affect the accuracy of such tracking. Researchers in the upcoming field of automated dietary monitoring (ADM) are attempting to track diet using electronic methods in an effort to mitigate this bias.

We attempt to replicate a previous algorithm that detects eating by tracking wrist motion electronically. The previous algorithm was evaluated on data collected from 43 subjects using an iPhone as the sensor. Periods of time are segmented first, and then classified using a naive Bayesian classifier. For replication, we describe the collection of the Clemson all-day data set (CAD), a free-living eating activity dataset containing 4,680 hours of wrist motion collected from 351 participants - the largest of its kind known to us. We learn that while different sensors are available to log wrist acceleration data, no unified convention exists, and this data must thus be transformed between conventions. We learn that the performance of the eating detection algorithm is affected due to changes in the sensors used to track wrist motion, increased variability in behavior due to a larger participant pool, and the ratio of eating to non-eating in the dataset.

We learn that commercially available acceleration sensors contain noise in their reported readings which affects wrist tracking specifically due to the low magnitude of wrist acceleration. Commercial accelerometers can have noise up to 0.06g which is acceptable in applications like automobile crash testing or pedestrian indoor navigation, but not in ones using wrist motion. We quantify linear acceleration noise in our free-living dataset. We explain sources of noise, a method to mitigate it, and also evaluate the effect of this noise on the eating detection algorithm.

By visualizing periods of eating in the collected dataset we learn that that people often conduct secondary activities while eating, such as walking, watching television, working, and doing household chores. These secondary activities cause wrist motions that obfuscate wrist motions associated with eating, which increases the difficulty of detecting periods of eating (meals). Subjects reported conducting secondary activities in 72% of meals. Analysis of wrist motion data revealed that the wrist was resting 12.8% of the time during self-reported meals, compared to only 6.8% of the time in a cafeteria dataset. Walking motion was found during 5.5% of the time during meals in free-living, compared to 0% in the cafeteria. Augmenting an eating detection classifier to include walking and resting detection improved the average per person accuracy from 74% to 77% on our free-living dataset (t[353]=7.86, p<0.001). This suggests that future data collections for eating activity detection should also collect detailed ground truth on secondary activities being conducted during eating.

Finally, learning from this data collection, we describe a convolutional neural network (CNN) to detect periods of eating by tracking wrist motion during everyday life. Eating uses hand-to-mouth gestures for ingestion, each of which lasts appx 1-5 sec. The novelty of our new approach is that we analyze a much longer window (0.5-15 min) that can contain other gestures related to eating, such as cutting or manipulating food, preparing foods for consumption, and resting between ingestion events. The context of these other gestures can improve the detection of periods of eating.

We found that accuracy at detecting eating increased by 15% in longer windows compared to shorter windows. Overall results on CAD were 89% detection of meals with 1.7 false positives for every true positive (FP/TP), and a time weighted accuracy of 80%.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.