ToF Based Wearable Sensing for Passive Food Intake Monitoring

Harshavardhan Sasikumar, Rashmi Wijesundara, Sarah Deemer, Xiaohui Yuan, Megan Wesling, Mahdi Pedram

Confirmation: I have read and agree with the IEEE BSN 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: Dietary Monitoring, Time-of-Flight (ToF) Sensor, RGB Image Masking, FOMO Object Detection, Eating Gesture Recognition, On-device processing, Health monitoring

TL;DR: A privacy-preserving chest-mounted wearable uses ToF sensing and on-device food detection and gesture recognition to monitor food intake.

Abstract: Diet plays a crucial role in preventing chronic diseases such as type 2 diabetes and heart disease. Most existing diet monitoring systems require manual input or raise privacy concerns by continuously recording video data that often captures the user’s face or the surrounding environment. In this paper, we present a chest-mounted wearable device that preserves user privacy while passively tracking dietary intake using a Time-of-Flight (ToF) sensor. Captured RGB images are masked using ToF depth data to isolate food items and eliminate background elements. A FOMO-based food detection model achieved an F1 score of 96\% and a mean Average Precision (mAP) of 74\% on masked images, outperforming its performance on unmasked RGB inputs. Also, ToF depth frames were used to build an eating gesture recognition model that achieved 88\% accuracy, indicating reliable identification of eating gestures. All models and image processing steps were executed on-device, demonstrating the feasibility of the system. This work presents a novel approach for real-time dietary monitoring that addresses both user privacy and food detection accuracy in a wearable health system.

Track: 1. Digital Health Solutions (i.e. sensors and algorithms) for diagnosis, progress, and self-management

NominateReviewer:

Harshavardhan Sasikumar
([e-mail])

Submission Number: 72