Contactless human activity recognition
The objective of this thesis is to design passive measurement and classification systems to recognize various human activities. Human Activity Recognition (HAR) can enhance a diverse range of human-centric applications in health care, smart homes, and security. Traditional solutions are based on wearable sensors and vision-based technologies; however, these solutions suffer from considerable limitations. The need for users to wear sensors for activity recognition is inconvenient and impractical for long periods of time. Contactless HAR systems have increased the abilities, practicality, and convenience of sensor based HAR systems. It allows for unobtrusive measurement and classification of sedentary behaviors in the workplace such as time spent sitting at desk, to intense physical activities such as at home exercises. This thesis provides contributions to contactless HAR methods and techniques using different sensing modalities. While activity recognition can be achieved with low cost ultrasound sensors, we show the beneficial impact of adopting radio frequency as a technological means of recognizing human activity. Specifically, WiFi signal analysis enables both macro and micro level activity classification. The ubiquity of WiFi infrastructures in home, university, work, and even outdoor environments makes WiFi the most convenient technology to produce valuable contributions in human activity recognition. The primary contributions of this work are: 1) Development of prototype hardware system, Echolocation Activity Detector (EAD), that achieves contactless activity recognition in office setting. 2) A low-cost system that employs WiFi monitoring of packets to provide a student engagement measure, and 3) Design of a non-invasive system that recognizes exercise activity and provides fine-grained repetition counting information of each exercise set using WiFi channel state information. Each of the contributions passively detects and classifies human activity in a contactless fashion without the need of wearable sensors.
Read
- In Collections
-
Electronic Theses & Dissertations
- Copyright Status
- In Copyright
- Material Type
-
Theses
- Authors
-
Harrington, Brandon Lamont
- Thesis Advisors
-
Biswas, Subir
- Committee Members
-
Sepulveda, Nelson
Mahapatra, Nihar
- Date
- 2020
- Subjects
-
Human activity recognition
- Program of Study
-
Electrical Engineering - Master of Science
- Degree Level
-
Masters
- Language
-
English
- Pages
- viii, 69 pages
- ISBN
-
9798645456863
- Permalink
- https://doi.org/doi:10.25335/4z16-8594