WAFFLE: A Wearable Approach to Bite Timing Estimation in Robot-Assisted Feeding

TL;DR

WAFFLE addresses the challenge of bite timing in robot-assisted feeding by leveraging a wearable sensing pipeline (IMU on glasses and a throat microphone) to predict the time until the next bite with a learned regression model. The system uses a threshold-based deployment to translate predictions into reactive robot actions, enabling natural, caregiver-like timing across individual and social dining and generalizing to different robots and user impairments. Across three phases, WAFFLE achieves superior or comparable accuracy to baselines, high user acceptance, and strong cognitive and physical workload advantages, demonstrating practical potential for real-world dining assistance. The approach preserves privacy, remains unobtrusive, and adapts to diverse contexts, suggesting significant impact for autonomous feeding devices and caregiver support.

Abstract

Millions of people around the world need assistance with feeding. Robotic feeding systems offer the potential to enhance autonomy and quality of life for individuals with impairments and reduce caregiver workload. However, their widespread adoption has been limited by technical challenges such as estimating bite timing, the appropriate moment for the robot to transfer food to a user's mouth. In this work, we introduce WAFFLE: Wearable Approach For Feeding with LEarned bite timing, a system that accurately predicts bite timing by leveraging wearable sensor data to be highly reactive to natural user cues such as head movements, chewing, and talking. We train a supervised regression model on bite timing data from 14 participants and incorporate a user-adjustable assertiveness threshold to convert predictions into proceed or stop commands. In a study with 15 participants without motor impairments with the Obi feeding robot, WAFFLE performs statistically on par with or better than baseline methods across measures of feeling of control, robot understanding, and workload, and is preferred by the majority of participants for both individual and social dining. We further demonstrate WAFFLE's generalizability in a study with 2 participants with motor impairments in their home environments using a Kinova 7DOF robot. Our findings support WAFFLE's effectiveness in enabling natural, reactive bite timing that generalizes across users, robot hardware, robot positioning, feeding trajectories, foods, and both individual and social dining contexts.

Figures (7)

• Figure 1: WAFFLE Deployment Pipeline. Features from 1 second windows of IMU and Throat Microphone data are inputted into our bite timing machine learning model which outputs $\hat{y}$, the predicted time (s) until the next bite. We use a user selected assertiveness threshold, $\tau$, to stop and progress the robot along the trajectory.
• Figure 2: Phase 2 participant responses to the question, "How was the bite timing of this method on average?" for both individual and social dining scenarios. Data points are additionally provided in Table \ref{['tab:bite_timing']}.
• Figure 3: Phase 2 and Phase 3 Individual and Social Dining Mean Likert Ratings with error bars representing the standard deviation. Higher scores indicate more favorable outcomes for all items. Asterisks denote statistically significant pairwise differences based on Wilcoxon signed-rank tests with Bonferroni correction ($p < 0.05$ = *, $p < 0.01$ = **, $p < 0.001$ = ***). Prior to pairwise comparisons, a non-parametric Friedman test was used to confirm overall differences among the three methods for each Likert item. Data is additionally included in Table \ref{['tab:individual_FI']}- \ref{['tab:social_W']}.
• Figure 4: Phase 2 preference rankings for the three methods. The Wearables method received the most 1st-place rankings across individual, social, and overall contexts.
• Figure A1: Study Setup for the Three Phases. A. Phase 1 Study Setup. In this study, we collect bite timing data from 14 participants with no motor impairments with the Obi robot feeding from the front of the participant. B. Phase 2 Study Setup. In this study, we compare our wearables approach, WAFFLE, to two baselines with 15 participants without motor impairments with the Obi robot feeding from the side of the participant. C. Phase 3 Study Setup. In this study, we evaluate our wearables approach with 2 participants with motor impairments with the FEAST robot jenamani2025feast feeding from the front of the participant.