Smart Wireless Sensing: From IoT to AIoT

ISBN-13: 9789811656606 / Angielski / Miękka / 2022 / 234 str.

Zheng Yang; Kun Qian; Chenshu Wu

Opis

Perception of human beings has evolved from natural biosensor to powerful sensors and sensor networks. In sensor networks, trillions of devices are interconnected and sense a broad spectrum of contexts for human beings, laying the foundation of Internet of Things (IoT). However, sensor technologies have several limitations relating to deployment cost and usability, which render them unacceptable for practical use. Consequently, the pursuit of convenience in human perception necessitates a wireless, sensorless and contactless sensing paradigm.Recent decades have witnessed rapid developments in wireless sensing technologies, in which sensors detect wireless signals (such as acoustic, light, and radio frequency) originally designed for data transmission or lighting. By analyzing the signal measurements on the receiver end, channel characteristics can be obtained to convey the sensing results. Currently, significant effort is being devoted to employing the ambient Wi-Fi, RFID, Bluetooth, ZigBee, and television signals for smart wireless sensing, eliminating the need for dedicated sensors and promoting the prospect of the Artificial Intelligence of Things (AIoT). This book provides a comprehensive and in-depth discussion of wireless sensing technologies. Specifically, with a particular focus on Wi-Fi-based sensing for understanding human behavior, it adopts a top-down approach to introduce three key topics: human detection, localization, and activity recognition. Presenting the latest advances in smart wireless sensing based on an extensive review of state-of-the-art research, it promotes the further development of this area and also contributes to interdisciplinary research.

Spis treści

Part I The Background

1 Human Sensing Modalities and Applications

1.1 What is Wireless Sensing

1.1.1 Definition

1.1.2 Wireless Signals

1.2 Characteristics of Wireless Sensing

1.3 Applications of Wireless Sensing

1.3.1 Smart Home

1.3.2 Security Surveillance

1.3.3 Vital/Biometrical Features Recognition

 

Part II Getting Started

2 Main Steps for Wireless Sensing

2.1 Data Collection

2.2 Signal Preprocessing

2.3 Feature Extraction

2.4 Model Training and Inference

 

Part III Detection: Passive Human Detection with Wireless Signals

3 The Background of Passive Human Detection

3.1 Motivation

3.2 Related Work

4 Passive Detection of Human with Dynamic Speed

4.1 Introduction

4.2 System Overview

4.3 Methodology

4.3.1 Data Processing

4.3.2 Feature Extraction

4.3.3 Motion Detection

4.3.4 Enhancement via Multiple Antennas

4.4 Experiments and Results

4.4.1 Experiment Setup

4.4.2 Performance Evaluation

4.5 Conclusions

5 Detection of Moving and Stationary Human with Wi-Fi

5.1 Introduction

5.2 Preliminary

5.3 System Design

5.3.1 Overview

5.3.2 Motion Inference Indicator

5.3.3 Moving Target Detection

5.4 Stationary Target Detection

5.4.1 Periodic Alterations from Breathing

5.4.2 Breathing Detection

5.4.3 Embracing Frequency Diversity

5.5 Experiments and Evaluation

5.5.1 Implementation

5.5.2 Performance

5.6 Discussions and Future Works

5.6.1 Monitoring Breathing Rate

5.6.2 Expanding Detection Coverage via Space Diversity

5.6.3 Multiple Target Detection

5.6.4 Extending to Through-Wall Detection

5.7 Conclusions

6 Omnidirectional Human Detection with Wi-Fi

6.1 Introduction

6.2 Preliminaries

6.2.1 The Omnidirectional Passive Human Detection Problem

6.2.2 Signal Power Features

6.3 Feature Extraction and Classification

6.3.1 Sensitivity to Human Presence

6.3.2 Resistance to Environmental Dynamics

6.3.3 Modeling CFR Amplitude Features

6.3.4 Signature Classification

6.4 Human Detection

6.5 Performance

6.5.1 Experiment Methodology

6.5.2 Static Detection Performance

6.5.3 The Impact of Window Size

6.5.4 Mobile Detection Performance

6.6 Conclusion

 

Part IV Localization: Passive Human Localization with Wireless Signals

7 The Background of Passive Human Localization

7.1 Motivation

7.2 Related Work

8 Human Localization via Velocity Monitoring with Wi-Fi

8.1 Introduction

8.2 Preliminary

8.2.1 Channel State Information

8.2.2 From CSI to PLCR

8.2.3 Challenges for Tracking

8.3 Modeling of CSI-Mobility

8.3.1 The Ideal Model

8.3.2 The Real Model

8.4 PLCR Extraction

8.4.1 CSI Preprocessing

8.4.2 PLCR Extraction Algorithm

8.4.3 PLCR Sign Identification

8.5 Tracking Velocity & Location

8.5.1 Movement Detection

8.5.2 Initial Location Estimation

8.5.3 Successive Tracking

8.5.4 Trace Refinement

8.6 Evaluation

8.6.1 Experiment Methodology

8.6.2 Overall Performance

8.6.3 Parameter Study

8.7 Conclusion

9 Human Localization with a Single Wi-Fi Link

9.1 Introduction

9.2 Overview

9.3 Motion in CSI

9.3.1 CSI-Motion Model

9.3.2 Joint Multiple Parameter Estimation

9.3.3 CSI Cleaning

9.4 Localization

9.4.1 Path Matching

9.4.2 Range Refinement

9.4.3 Localization Model

9.5 Evaluation

9.5.1 Experiment Methodology

9.5.2 System Performance

9.5.3 Parameter Study

9.6 Discussion

9.7 Conclusion

 

Part V Recognition: Passive Human Activity Recognition with Wireless Signals

10 The Background of Passive Human Activity Recognition

10.1 Motivation

10.2 Related Work

11 Moving Direction Estimation with Wi-Fi

11.1 Introduction

11.2 Overview

11.3 Doppler Effect in Wi-Fi

11.3.1 Doppler Effect

11.3.2 Doppler Effect in CSI

11.3.3 Doppler Effect by Multiple Antennas

11.3.4 Extraction of Doppler Effect

11.4 Motion Recognition

11.4.1 Player Reaction in Doppler Effect

11.4.2 Motion Recognition Workflow

11.5 Evaluation

11.5.1 Experiment Methodology

11.5.2 Performance

11.5.3 Parameter Study

11.6 Limitations and Discussions

11.6.1 Wireless Sensing Systems

11.6.2 Wi-Fi-based Gesture Sensing Systems

11.6.3 Interfaces for Exergames

11.7 Conclusion

12 Environment-Independent Gesture Recognition

12.1 Introduction

12.2 Motivation

12.2.1 Primitive Features without Cross-Domain Capability

12.2.2 Cross-Domain Motion Features for Coarse Tracking

12.2.3 Latent Features from Cross-Domain Learning Methods

12.2.4 Lessons Learned

12.3 Overview

12.4 Body-Coordinate Velocity Profile

12.4.1 Doppler Representation of CSI

12.4.2 From DFS to BVP

12.4.3 BVP Estimation

12.4.4 Location and Orientation Prerequisites

12.5 Recognition Mechanism

12.5.1 BVP Normalization

12.5.2 Spatial Feature Extraction

12.5.3 Temporal Modeling

12.6 Evaluation

12.6.1 Experiment Methodology

12.6.2 Overall Accuracy

12.6.3 Cross-Domain Evaluation

12.6.4 Method Comparison

12.6.5 Parameter Study

12.7 Discussions

12.7.1 User Height

12.7.2 Number of Wi-Fi Links for Gesture Recognition

12.7.3 Applications Beyond Gesture Recognition

12.8 Conclusion

13 Human Gait Recognition with Wi-Fi

13.1 Introduction

13.2 Motivation

13.2.1 Immune to Trajectory and Speed Variance

13.2.2 Reducing Training Data for Newcomers

13.2.3 Lessons Learned

13.3 System Design

13.3.1 GBVP Extraction

13.3.2 Recognition Mechanism

13.4 Evaluation

13.4.1 Experimental Methodology

13.4.2 Overall Performance

13.4.3 Generalizability Evaluation

13.4.4 Parameter Study

13.5 Conclusion

 

Part VI Conclusions

14 Research Summary and Open Issues

14.1 Research Summary

14.2 Open Issues

 

O autorze

Zheng Yang is an associate professor at the School of Software and BNRist, Tsinghua University. He holds a BE degree from Tsinghua University, and a PhD degree from Hong Kong University of Science and Technology. His research interests include Internet of Things, mobile computing, pervasive computing, industrial internet, smart city, etc. He is the author and co-author of 3 books and over 100 papers published in leading journals and conferences. Zheng received the China National Natural Science Award (2011). He is a senior member of IEEE and a member of ACM.

Kun Qian is a post-doctoral researcher in the Department of Electrical and Computer Engineering, University of California San Diego. He received his Ph.D in 2019 at the School of Software, Tsinghua University. He received his B.E. in 2014 in Software Engineering from School of Software, Tsinghua University. His research interests include mobile computing and wireless sensing, etc. He has published over 20 papers in competitive conferences and journals.

Chenshu Wu is an assistant professor at the University of Hong Kong. He is also the Chief Scientist at Origin Wireless Inc. His research focuses on wireless AIoT systems at the intersection of wireless sensing, ubiquitous computing, and the Internet of Things. He has published two books, over 60 papers in prestigious conferences and journals, and over 40 patents. His research has been commercialized as products, including LinkSys Aware that won the CES 2020 Innovation Award, HEX Home that won CES 2021 Innovation Award, and Origin Health Remote Patient Monitoring that won CES 2021 Best of Innovation Award. He holds BS and PhD degrees in Computer Science both from Tsinghua University.

Yi Zhang is currently working toward his PhD degree at the School of Software in Tsinghua University. Prior to that, he received his BE degree from the School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, in 2017. His research interests include wireless sensing, mobile computing, and machine learning. He is the author and co-author of over 6 papers published in leading journals and conferences.