SCH: Wearable Sensing and Visual Analytics to Estimate Receptivity to Just-In-Time Interventions for Eating Behavior

不良饮食是可预防的死亡和疾病的主要原因，也是可预防的医疗费用的主要原因。 美国。尽管遵循健康的饮食模式很重要，但大多数美国成年人并不 符合国家饮食指南，并且超重或肥胖。迫切需要 “及时”(JIT)干预，在饮食和饮食行为发生时改善它们。为了最大限度地发挥影响力， JIT干预应仅在个体易于接受的情况下提供，尤其是在饮食质量 很穷。然而，食物环境和饮食行为的哪些方面会对饮食产生影响 摄入量和质量尚不清楚，它们与JIT干预接受性的关系也未被探索。这 需要收集和分析日常生活背景下关于一个人的饮食的几乎连续的数据， 行为实际发生的地方，这对研究人员来说是非常具有挑战性的，对 参与者。配备新型计算方法的可穿戴传感器技术的进展 可以提供一种途径来捕捉和分析进食中的各种暴露和模式 环境来填补这一空白。这项提议的总体目标是创建一个综合的系统 用于发现与饮食有关的食物环境暴露的可穿戴传感器和计算方法 影响JIT干预接受度的质量。在这一愿景的推动下，本研究的目标 包括：1)开发用于隐私保护压缩新型边缘计算硬件和软件 图像捕获和传输，2)开发新的协作压缩和分析， 无人监督的持续学习来理解饮食行为，3)确定感知到的方面是否 进食期间的环境背景与饮食质量和对JIT干预的接受度有关， 尤其是当饮食质量很差的时候。该项目是一个合作努力，结合了以下方面的专业知识 可穿戴电子产品、图像处理、饮食模式和行为科学。

Poor diet is a leading cause of preventable death and diseases, as well as preventable healthcare costs in the United States. Despite the importance of following a healthy dietary pattern, most U.S. adults do not meet national dietary guidelines and are either overweight or obese. There is a critical need for "just-in-time" (JIT) interventions to improve diet and eating behaviors as they occur. To maximize impact, JIT interventions should only be delivered when an individual is receptive, particularly when dietary quality is poor. However, which aspects of the food environment and dietary behavior have influence on dietary intake and quality are unknown, and how they relate to JIT intervention receptivity is unexplored. This would require collecting and analyzing near-continuous data about one's diet in the context of daily life, where behavior actually occurs, which is very challenging for researchers and burdensome for participants. Advances in wearable sensor technologies, equipped with novel computational methods could provide a pathway to capture and analyze the various exposures and patterns in the eating environment to fill this gap. The overall objective of this proposal is to create an integrated system of wearable sensor and computational methods to discover food environment exposures related to dietary quality that influence JIT intervention receptivity. Motivated by this vision, the objectives of this research include: 1) develop novel edge computing hardware and software for privacy-preserving compressive image capture and transmission, 2) develop new collaborative compression and analytics together with unsupervised continual learning to understand eating behavior, 3) determine whether sensed aspects of the environmental context during eating relate to dietary quality and receptivity to JIT interventions, particularly when the dietary quality is poor. The project is a collaborative effort combining expertise in wearable electronics, image processing, dietary patterns, and behavioral science.