My research focuses on enabling contactless health sensing using high-frequency millimeter-wave wireless devices to provide an effective alternative or acts as complementary to the existing systems to enable fine-grained health monitoring.
Human Posture Estimation
Our work designs and validates a learning-based mmWave imaging system that can generate high-resolution human silhouettes using COTS devices to estimate human postures. Vision-based systems, such as optical cameras, IRs, and LiDARs, enable contactless human posture estimation. However, they do not perform well in low light and low visibility conditions. Cameras, in particular, pose privacy concerns due to their true-color image capture, making them less desirable for use in private spaces such as homes or offices. In contrast, 5G-and-beyond devices utilizing high-frequency mmWave wireless signals are not hindered by low light or visibility and can facilitate imaging through occlusions. Our experimental results from 10 volunteers demonstrate that our work delivers high-resolution silhouettes and accurate body poses on par with an existing vision-based system.
Read our articles here: [IMWUT/UbiComp’22] [UbiComp-ISWC'21] [MobiSys'22] [News UofSC]
Contactless Sleep Monitoring
Our work expands the sensing capabilities of 5G-and-beyond devices to enable fine-grained sleep monitoring. MmWave technology in 5G-and-beyond devices, with its higher resolution compared to Wi-Fi,can provide detailed insights into body posture even in dark conditions and under the blanket without requiring extra hardware. Identifying sleep postures directly from the mmWave signals, however, is challenging. We have designed a height-agnostic sytem that provides fine-grained information to infer sleep postures in the form of 3D joint locations, and detects key sleep events.
Read our articles here: [ACM TIOT] [IEEE INFOCOM'23] [IEEE INFOCOM'23] [Demo]
Vital Sign Monitoring
At-home Spirometry
Our research presents a low-barrier means to perform at-home spirometry tests using 5G smart devices. Our system works like regular spirometers, where a user forcibly exhales onto a device; but instead of relying on special-purpose hardware, it leverages built-in millimeter-wave technology in general-purpose, ubiquitous mobile devices. Our system analyzes the tiny vibrations created by the airflow on the device surface and combines wireless signal processing with deep learning to enable a software-only spirometry solution. From empirical evaluations, we find that mmFlow can predict the spirometry indicators with performance comparable to in-clinic spirometers.