Gesture Recognition with Wi-Fi Sensing in Smart Homes


Krunal Shah

Oral Defence Date: 

Wednesday, July 24, 2019 - 10:00


TH 434


Assistant Prof. Hao Yue and Prof. James Wong


With the fast growth of smart appliances and smart devices at home, A Smart Homes have received increased attention and have been translating from a vision into reality. In Smart Homes, we need to track and recognize the activities and gestures of residents to intelligently control home appliances and achieve comfort, convenience, security, and energy efficiency. A Conventional sensing methodology for Smart Homes is labor-intensive and complicated for practical deployment. In this project, we propose a novel alternative sensing method for Smart Homes, which utilizes the prevalent Wi-Fi devices and networks at home to detect and recognize human gestures. Specifically, we build a machine learning model to examine and extract the unique patterns exhibited in the Channel State Information available in the received Wi-Fi signals at wireless devices and use it to sense and identify human gestures in an indoor environment. The awareness of human dynamics is critically important in designing management services, especially for care services for seniors. Human monitoring aims to increase personal autonomy of elderly and impaired people by means of ambient-assisted living(AAL) to detect a home accident, long term behavior analysis, telecare. Traditional approaches are often device-based approaches which require installing dedicated devices or sensor which can be an overhead. However, the approach that we have used is a low-cost Wi-Fi device to infer comprehensive activity-related information leveraging Channel State Information (CSI). In this approach, we first collect the data through CSI of existing Wi-Fi devices.

Krunal Shah

machine learning, Wi-Fi Sensing, Auto-encoder, Matlab, Channel State Information(CSI), Human Monitoring, Human Dynamics, Smart Home, Gesture Recognition