Health-risk behaviors such as overeating and smoking have a profound impact on public health, making their monitoring and mitigation critical. Wearable RGB-Thermal cameras are being employed to monitor these behaviors by capturing hand-to-mouth (HTM) gestures, which are central to them. However, detection models relying on single modalities—either RGB or thermal—often struggle to accurately distinguish these confounding gestures due to inherent sensor limitations, such as sensitivity to lighting conditions or thermal occlusions. We present a family of fusion models that integrate RGB and thermal video data using early-, decision- , and a novel mid-fusion architecture, RGB-Thermal Fusion Video Network (RTFVNet), designed to enhance the recognition of HTM gestures associated with eating and smoking. Our evaluation shows that while decision fusion achieves the highest F1-score of 88% (0.44 TFLOPs), RTFVNet offers an optimal balance between performance (85%) and complexity (0.37 TFLOPs) for gesture classification of eating, smoking, and non-gesture activities.

Excessive ultraviolet (UV) exposure is the principal driver of melanoma, yet at-risk individuals seldom receive timely, context-aware cues to apply protection. Existing wrist-worn UV monitors often struggle to recognize timely outdoor exposure because UV readings vary sharply with wrist orientation and sensor angle. To address this gap, we developed a wrist-watch form-factor device that embeds an AS7331 UV photodiode beneath a hemispherical polytetrafluoroethylene (PTFE) dome, which diffuses incident light and stabilizes the sensor's angular response. Ten participants wore the device during routine daily activities, generating more than 93k datapoints of annotated indoor-outdoor data. We implemented an on-device logistic-regression classifier, integrating UVA, UVB, and inertial features to distinguish indoor from outdoor contexts. Under leave-one-participant-out cross-validation, the PTFE-dome watch achieved 94% accuracy and a weighted F1 score of 0.95, outperforming both a flat-window GUVA-S12SD sensor (71% accuracy, F1 = 0.72) and a cylindrical-PTFE enclosure (84% accuracy, F1 = 0.85). By coupling a compact PTFE dome with on-device machine learning (ML), our system delivers reliable, on-wrist UV context sensing and paves the way for unobtrusive, personalized interventions to reduce cumulative UV exposure.

Overeating is a key contributor to obesity, yet identifying and characterizing its underlying causes remains challenging. While prior research has leveraged Ecological Momentary Assessment (EMA) to capture psychological and contextual factors in real-time, few studies have integrated EMA with passive sensing to uncover fine-grained, individualized consumption behaviors. In this work, we present a multimodal framework combining psychological and contextual data from a custom-built EMA app with validated camera-derived meal microstructure features from a neck-worn activity-oriented wearable camera. Across 41 participants, the camera captured 6,343 hours of footage over 312 days, yielding annotated bites, chews, meal start/end times, and dietitian-confirmed caloric intake. Using supervised contrastive learning, we generated meal-level representations, projected them using UMAP, and applied k-means clustering to identify behavioral phenotypes. We then conducted a z-score analysis to highlight features most distinctive to each cluster. Among the eight discovered groups, three consistently showed high purity for overeating meals (average purity$=0.99$), revealing nuanced, data-driven overeating phenotypes that may inform targeted intervention strategies.