NeckSense: A Multi-Sensor Necklace for Detecting Eating Activities in Free-Living Conditions
summary
A multi-sensor necklace that uses a proximity sensor, ambient light, and lean forward angle plus an energy proxy to detect chewing sequences and cluster them into eating episodes, running longest periodic subsequence peak finding and XGBoost classification on >470 hours of in-the-wild data. Tested on 20 participants (11 with obesity, 9 without), achieves 81.6% F1 in an exploratory study and 77.1% F1 in free-living, with ~15.8 hours battery life and released dataset and code.
AI-written summary · about
chapters (5)
- 0:00Introduction to NextSense
- 1:34Sensor system design
- 2:51Algorithm and pipeline
- 4:36Study results and findings
- 5:21Conclusion and future impact
related talks
06mOne More Bite? Inferring Food Consumption Level of College Students Using Smartphone Sensing and ...
01h 13mHealth Monitoring with Wireless Sensors and Machine Learning
08mSenseCollect: We Need Efficient Ways to Collect On-body Sensor-based Human Activity Data!
06mDetecting Stressful Social Interactions Using Wearable Physiological and Inertial Sensors
06mOpiTrack: A Wearable-based Clinical Opioid Use Tracker with Temporal Convolutional Attention ...
06mmTeeth: Identifying Brushing Teeth Surfaces UsingWrist-Worn Inertial Sensors
09mAn Activity Recognition System for Taking Medicine Using In-The-Wild Data to Promote Medication ...
44mQuantified Self: Analyzing the Big Data of our Daily Life