PROMOTING WIFI-BASED MOTION SENSING USING CHANNEL STATE INFORMATION

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "PROMOTING WIFI-BASED MOTION SENSING USING CHANNEL STATE INFORMATION"

By

Miss Yuxi WANG


Abstract

Motion sensing, which studies the changes of people's location and posture, has 
been applied for various applications, such as location-based services, 
household appliance control, and motion sensing games. Motion sensing systems 
can be classified into two categories for sensing purposes: location-related 
and posture-related. Location-related motion sensing systems detect the 
instantaneous location or the moving trajectories of human, while 
posture-related motion sensing approaches sense the posture variation and the 
human movements. Various techniques have been utilized to achieve 
motion sensing among which WiFi begins to receive more attention from academia 
and industry in the past decades due to the wide deployment and low cost of 
WiFi infrastructure. WiFi has been first adopted for indoor location-based 
services and then widely applied for human activity and gesture recognition. In 
this thesis, we follow this line of research and propose three WiFi-based 
motion sensing systems to enhance location-related and posture-related motion 
sensing.

We take advantage of the WiFi physical layer channel state information and 
propose motion sensing systems for both sensing purposes. For location-related 
motion sensing, we propose WiShape which can sense the shape of the moving 
trajectory. WiShape studies the WiFi signal variation pattern of different 
trajectory shapes and shows a potential in corner shape detection. Then, we pay 
more attention to posture-related motion sensing and propose two device-free 
passive sensing systems, namely WiFall and WiWrite. WiFall establishes the 
relationship between WiFi signal variation and human activities to achieve 
precise fall detection. Our improvement of WiFall allows it to directly 
classify other daily activities indoors. We then study human limb posture and 
propose WiWrite, a device-free finger writing system. WiWrite detects the 
basic strokes decomposed from upper-case English letters by analyzing WiFi 
signal variation, and constructs English alphabets and words based on the 
strokes to achieve text entry.


Date:			Monday, 24 July 2017

Time:			1:00pm - 3:00pm

Venue:			Room 2130C
 			Lifts 19

Chairman:		Prof. Chi-Ying Tsui (ECE)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Prof. Lei Chen (Supervisor)
 			Prof. Shing-Chi Cheung
 			Prof. Qiong Luo
 			Prof. Jingshen Wu (MAE)
 			Prof. Jiannong Cao (Computing, PolyU)


**** ALL are Welcome ****