WiFi based device-free human activity recognition

• Main Author: Zhu, Zheyu
• Other Authors: Xie Lihua
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/141049
• Institution: Nanyang Technological University

Nowadays, we are in an era of information explosion. 5G technology and artificial intelligence are in the center of this technological revolution. The development of wireless devices and semiconductor devices has laid the foundation for the implementation of human behavior recognition technology. As one of the hot research topics in the Internet of Things technology, behavior recognition technology can provide convenience for many people and has strong practical application value. Application examples like gesture recognition, orientation recognition, motion tracking, smart home, security, surveillance, somatosensory games, virtual reality and so on, which are all of great prospects, high research significance and economic value. Compared with the traditional human behavior recognition technology, device-free HBR technology has the advantages that the user does not need to wear a device, signal can be transmitted through wall, wide-coverage, devices can work at night without any light, no dead ends, and protects user privacy better. Because the technology is based on WIFI equipment, it is very practical to apply these technologies by households in various fields. This report investigates a variety of methods for realizing human behavior recognition, compares three traditional methods with WIFI-based behavior recognition, and analyzes their achievability, advantages and disadvantages. Then, in terms of wireless detection, RSSI and CSI are compared, providing detailed reasons for using CSI and investigating the detection method of human fall and other movements based on CSI. This is followed by an introduction to some theory of MIMO and CSI. And this report shows our work that we discuss some existing CNN and machine learning development. Then deep CORAL is introduced which is the key of our algorithm. In terms of hardware, our platform can receive 114 subcarriers compared with other traditional CSI tool, and can provide better accuracy rate. Finally, this project shows the process of applying the neural network method on the HBR system based on WIFI to classify six behaviors in four different environments and how data are collected and experimental tests are carried out. And our system finally can get an accuracy of 89.7%, when the interference is moderate.