CAREER: Ubiquitous Sensing Using Computational Light

感知和检测人体运动的能力对于数据驱动的移动卫生系统的发展至关重要。它可以帮助检测疾病，促进行为改变，培养健康的生活方式。现有的传感技术要么需要用户不断佩戴或随身携带可能很笨重的设备，要么容易受到电磁干扰，要么存在严重的隐私风险，涉及敏感数据和图像的泄露。这个项目采用了一种完全不同的方法来解决这些问题。它利用无处不在的光作为一种低成本、不显眼、准确的传感媒介，能够同时感知人和周围环境。提出的愿景“LightSense”由天花板上现成的LED灯和散布在环境中的一些低成本光电二极管传感器组成。光电二极管被动捕捉人体产生的光阻塞，并实时重建细粒度的用户行为。LightSense利用光线将一个空间变成一个认知空间，它可以识别我们的存在，感知我们的行为，如姿势和高水平的活动，同时监测我们的健康状况指标，如压力水平。LightSense通过可见光通信（VLC）将可见光转换为计算光。它包含以下新颖的系统和算法设计：1)采用LED面板和稀疏光电二极管的新型VLC网络架构，以简化系统部署；2)算法和系统设计，以从密集的led中分离光线，优化光电二极管的放置，并克服其他物体（例如，家具，其他用户）的阻挡；3)一种新的VLC原语，即使在极低的光照条件下也能保持光通信和传感；4)学习算法来推断身体活动，推导运动特征，监测心理状态。LightSense将使用实际规模的测试平台和用户研究进行评估。该项目的成果将为定义一个新的研究空间（可见光传感）奠定基础，并将对促进创新交互设计和实现新型精确健康监测产生深远影响。

The ability to sense and detect human movement is critical to the development of data-driven mobile health systems. It can help detect disease and foster behavioral changes to cultivate healthy lifestyles. Existing sensing technologies either require users to constantly wear or carry on-body potentially cumbersome devices, are vulnerable to electromagnetic interference, or present severe privacy risks involving leaking of sensitive data and images. This project takes a entirely different approach to addressing these issues. It exploits the use of ubiquitous light as a low-cost, unobtrusive, and accurate sensing medium capable of simultaneously sensing people and their surrounding context. The proposed vision " LightSense " consists of off-the-shelf LED lights on the ceiling and a few low-cost photodiode sensors sprinkled in the environment. The photodiodes passively capture light blockage created by the human body and reconstruct fine-grained user behaviors in real time. LightSense leverages light to turn a space into a cognitive space, which recognizes our presence, senses our behaviors such as postures and high-level activities while monitoring our health status indicators such as levels of stress. LightSense is empowered by Visible Light Communication (VLC) that turns the visible light into computational light. It contains the following novel systems and algorithmic designs: 1) a novel VLC network architecture with LED panels and sparse photodiodes to ease system deployment; 2) algorithmic and systems designs to separate light rays from dense LEDs, optimize the placement of photodiodes, and overcome the blockage of other objects (e.g., furniture, other users); 3) a new VLC primitive that allows light communication and sensing to be sustained even under extremely low light conditions; and 4) learning algorithms to infer physical activities, derive movement characteristics, and monitor psychological state. LightSense will be evaluated using real-scale testbeds and user studies. Results from this project will establish the foundational pieces to define a new research space (visible light sensing), and will generate far-reaching impact on promoting innovative interaction designs and enabling new types of precise health monitoring.