CubeSense++: Smart Environment Sensing with Interaction-Powered Corner Reflector Mechanisms

要旨

Smart environment sensing provides valuable contextual informa- tion by detecting occurrences of events such as human activities and changes of object status, enabling computers to collect per- sonal and environmental informatics to perform timely responses to user’s needs. Conventional approaches either rely on tags that re- quire batteries and frequent maintenance, or have limited detection capabilities bounded by only a few coarsely predefined activities. In response, this paper explores corner reflector mechanisms that encode user interactions with everyday objects into structured responses to millimeter wave radar, which has the potential for integration into smart environment entities such as speakers, light bulbs, thermostats, and autonomous vehicles. We presented the design space of 3D printed reflectors and gear mechanisms, which are low-cost, durable, battery-free, and can retrofit to a wide array of objects. These mechanisms convert the kinetic energy from user interactions into rotational motions of corner reflectors which we computationally designed with a genetic algorithm. We built an end-to-end radar detection pipeline to recognize fine-grained activity information such as state, direction, rate, count, and usage based on the characteristics of radar responses. We conducted stud- ies for multiple instrumented objects in both indoor and outdoor environments, with promising results demonstrating the feasibility of the proposed approach.

著者
Xiaoying Yang
University of California, Los Angeles, Los Angeles, California, United States
Jacob Sayono
University of California, Los Angeles, Los Angeles, California, United States
Yang Zhang
University of California, Los Angeles, Los Angeles, California, United States
論文URL

https://doi.org/10.1145/3586183.3606744

動画

会議: UIST 2023

ACM Symposium on User Interface Software and Technology

セッション: Sensing Sorcery: Novel Sensing Techniques and Systems

Venetian Room
6 件の発表
2023-11-01 01:00:00
2023-11-01 02:20:00