In this project, we propose to bring IoT to active noise cancellation by combining wireless communication with acoustics. The core idea is to place an IoT device in the environment that listens to ambient sounds and forwards the sound over its wireless radio. Since wireless signals travel much faster than sound, our ear-device receives the sound in advance of its actual arrival. This serves as a glimpse into the future, that we call lookahead, and proves crucial for real-time noise cancellation.
Paper in SIGCOMM 2018.
We have shown that by utilizing hardware nonlinearities, inaudible signals (at ultrasound frequencies) can be designed to be audible to microphones. This can empower an adversary to stand on the road and silently control Amazon Echo and Google Home-like devices in people’s homes. We further push the range of this attack to be as far as 25 feet, limited by the power of our amplifier. We also develop a defense against this class of voice attacks that exploit non-linearity, which only require software changes at microphone.
Paper in NSDI 2018.
We have recently started to explore the possibility of bringing IoT to sports analytics, a thriving industry in which motion patterns of balls and players are being analyzed for coaching and predictions. We intend to decompose and track the motion of the ball, racquet, arm and body, with inexpensive IoT sensors and radios. The core problem pertains to statistical decomposition, motion tracking, signal processing and sensor fusion. We have developed a first system to track the 3D trajectory and spin parameters of a Cricket ball, with IMU sensors and UWB radios embedded in the ball.
Paper in NSDI 2017.
In this project, we explore the possibility of tracking the entire arm posture using only the IMU sensors on a smartwatch. We design ArmTrak, which fuses the data from IMU sensors and observations from human kinematics into a hidden Markov model to continuously estimate the 3D arm posture. We hope with some additional work, ArmTrak become a useful underlay to various practical applications.
Paper in ACM MobiSys 2016.
In this project, we explore an opportunity for automatic semantic localization – the presence of a website corresponding to each physical store. We propose to correlate the information seen in a physical store with that found in websites of the stores around that location, to recognize that store. Specifically, we assume a repository of crowdsourced WiFi-tagged pictures from different stores. By correlating words inside the pictures, against words extracted from store websites, our proposed system can automatically label clusters of pictures, and the corresponding WiFi APs, with the store name. Later, when a user enters a store, her smartphone can scan the WiFi APs and consult a lookup table to recognize the store she is in.
Paper in WPA 2015.
This project aimed at constructing the 3D map inside indoor environments by use of mobile phones. We utilized our accurate indoor localization techniques to boost depth point matching and multiple images based 3D reconstruction. Moreover, we achieved localization-based real-time rendering on Windows Phone devices.
Demo at Microsoft Techfest 2014.
This project aimed at integrating devices using virtualization technology, i.e., making phone and computer work together to extend capabilities and create rich scenarios using Virtual Device Driver technology. We developed several virtual drivers that seamlessly extended computer’s features using sensors on mobile devices such as GPS (Bing Map), accelerometer (Gaming), touchscreen (PowerPoint), etc. I developed a virtual display driver and incorporated a touch driver in order to turn mobile devices, such as a tablet, into an extra touchscreen of PC.
Demo at Microsoft Techfest 2014.
GPS works well for localization in an outdoor environment, but the signal is too weak to penetrate roofs and walls, making it useless for indoor localization. Therefore, we proposed a WiFi-based positioning system to localize a mobile user inside indoor environments. Our localization technique was based on measuring the received signal strength (RSS), and we boosted localization accuracy by deriving and applying models that were adapted to local environmental properties. I was also designing semi-supervised learning algorithms that leverage both unlabeled data and user feedbacks to reduce training effort. We had achieved average distance error less that 3m in our building.
I applied for an internship at Optical Network Lab in our university, where we mainly studied service provisioning in OFDM-based optical networks. In the project I analyzed routing, modulation and spectrum assignment under the advanced reservation traffic model, where requests specify their start and holding time in advance (such as video conferencing and grid applications), and proposed several efficient heuristics and theoretical analysis for the problem. Some of my results were accepted for ECOC 2013, in which I developed solutions for dynamic advanced reservation multicast by considering routing and spectrum assignment jointly to reduce blocking rate.
Paper in ECOC 2013.