Body Sensor Networking, Design and Algorithms
Description
In Body Sensor Networking, Design, and Algorithms, professionals in the field of Biomedical Engineering and e-health get an in-depth look at advancements, changes, and developments. When it comes to advances in the industry, the text looks at cooperative networks, noninvasive and implantable sensor microelectronics, wireless sensor networks, platforms, and optimization--to name a few.
Each chapter provides essential information needed to understand the current landscape of technology and mechanical developments. It covers subjects including Physiological Sensors, Sleep Stage Classification, Contactless Monitoring, and much more.
Among the many topics covered, the text also includes additions such as:
* Over 120 figures, charts, and tables to assist with the understanding of complex topics
* Design examples and detailed experimental works
* A companion website featuring MATLAB and selected data sets
Additionally, readers will learn about wearable and implantable devices, invasive and noninvasive monitoring, biocompatibility, and the tools and platforms for long-term, low-power deployment of wireless communications. It's an essential resource for understanding the applications and practical implementation of BSN when it comes to elderly care, how to manage patients with chronic illnesses and diseases, and use cases for rehabilitation.
More details
Other editions
Additional editions
Persons
Saeid Sanei is a Professor of Biomedical Signal Processing and Machine Learning at Nottingham Trent University and a Visiting Professor to Imperial College London, in the United Kingdom. His major contributions in advanced signal processing techniques such as tensor factorization, cooperative networking, compressive sensing, statistical signal processing, and subspace analysis have applications in physiological signal processing and sensor networks as explored in his three published monograms and over 400 publications.
Delaram Jarchi is currently a Lecturer at Essex University. She has been working intensively on sensor networks design and algorithms levels. Her research is focused on designing new algorithms and validation of commercial wearable sensors for robust estimation of physiological parameters such as heart rate, respiratory rate and blood oxygen saturation levels in very unobtrusive ways. She is a senior member of IEEE since 2018.
Anthony G. Constantinides is a Professor at Imperial College of London UK. He is an IEEE acknowledged pioneer in signal processing with research interests that span a wide range of applications of the area. Amongst these and relevant to the present book are included topics such as data analytics, acquisition, sensing, transmission, and compression.
Content
Preface xiii
About the Companion Website xv
1 Introduction 1
1.1 History of Wearable Technology 1
1.2 Introduction to BSN Technology 2
1.3 BSN Architecture 7
1.4 Layout of the Book 10
References 11
2 Physical, Physiological, Biological, and Behavioural States of the Human Body 17
2.1 Introduction 17
2.2 Physical State of the Human Body 17
2.3 Physiological State of Human Body 19
2.4 Biological State of Human Body 23
2.5 Psychological and Behavioural State of the Human Body 24
2.6 Summary and Conclusions 30
References 31
3 Physical, Physiological, and Biological Measurements 35
3.1 Introduction 35
3.2 Wearable Technology for Gait Monitoring 35
3.2.1 Accelerometer and Its Application to Gait Monitoring 36
3.2.1.1 How Accelerometers Operate 37
3.2.1.2 Accelerometers in Practice 39
3.2.2 Gyroscope and IMU 40
3.2.3 Force Plates 41
3.2.4 Goniometer 41
3.2.5 Electromyography 41
3.2.6 Sensing Fabric 42
3.3 Physiological Sensors 42
3.3.1 Multichannel Measurement of the Nerves Electric Potentials 42
3.3.2 Other Sensors 45
3.4 Biological Sensors 48
3.4.1 The Structures of Biological Sensors - The Principles 48
3.4.2 Emerging Biosensor Technologies 51
3.5 Conclusions 51
References 53
4 Ambulatory and Popular Sensor Measurements 59
4.1 Introduction 59
4.2 Heart Rate 59
4.2.1 HR During Physical Exercise 60
4.3 Respiration 62
4.4 Blood Oxygen Saturation Level 67
4.5 Blood Pressure 70
4.5.1 Cuffless Blood Pressure Measurement 71
4.6 Blood Glucose 72
4.7 Body Temperature 73
4.8 Commercial Sensors 74
4.9 Conclusions 75
References 76
5 Polysomnography and Sleep Analysis 83
5.1 Introduction 83
5.2 Polysomnography 84
5.3 Sleep Stage Classification 85
5.3.1 Sleep Stages 85
5.3.2 EEG-Based Classification of Sleep Stages 86
5.3.2.1 Time Domain Features 86
5.3.2.2 Frequency Domain Features 87
5.3.2.3 Time-frequency Domain Features 87
5.3.2.4 Short-time Fourier Transform 88
5.3.2.5 Wavelet Transform 88
5.3.2.6 Matching Pursuit 88
5.3.2.7 Empirical Mode Decomposition 89
5.3.2.8 Nonlinear Features 89
5.3.3 Classification Techniques 90
5.3.3.1 Using Neural Networks 90
5.3.3.2 Application of CNNs 92
5.3.4 Sleep Stage Scoring Using CNN 94
5.4 Monitoring Movements and Body Position During Sleep 96
5.5 Conclusions 99
References 100
6 Noninvasive, Intrusive, and Nonintrusive Measurements 107
6.1 Introduction 107
6.2 Noninvasive Monitoring 107
6.3 Contactless Monitoring 109
6.3.1 Remote Photoplethysmography 109
6.3.1.1 Derivation of Remote PPG 110
6.3.2 Spectral Analysis Using Autoregressive Modelling 111
6.3.3 Estimation of Physiological Parameters Using Remote PPG 114
6.3.3.1 Heart Rate Estimation 114
6.3.3.2 Respiratory Rate Estimation 116
6.3.3.3 Blood Oxygen Saturation Level Estimation 117
6.3.3.4 Pulse Transmit Time Estimation 118
6.3.3.5 Video Pre-processing 119
6.3.3.6 Selection of Regions of Interest 120
6.3.3.7 Derivation of the rPPG Signal 120
6.3.3.8 Processing rPPG Signals 120
6.3.3.9 Calculation of rPTT/dPTT 121
6.4 Implantable Sensor Systems 122
6.5 Conclusions 123
References 124
7 Single and Multiple Sensor Networking for Gait Analysis 129
7.1 Introduction 129
7.2 Gait Events and Parameters 129
7.2.1 Gait Events 129
7.2.2 Gait Parameters 130
7.2.2.1 Temporal Gait Parameters 130
7.2.2.2 Spatial Gait Parameters 132
7.2.2.3 Kinetic Gait Parameters 133
7.2.2.4 Kinematic Gait Parameters 133
7.3 Standard Gait Measurement Systems 135
7.3.1 Foot Plantar Pressure System 135
7.3.2 Force-plate Measurement System 135
7.3.3 Optical Motion Capture Systems 137
7.3.4 Microsoft Kinect Image and Depth Sensors 138
7.4 Wearable Sensors for Gait Analysis 140
7.4.1 Single Sensor Platforms 140
7.4.2 Multiple Sensor Platforms 141
7.5 Gait Analysis Algorithms Based on Accelerometer/Gyroscope 143
7.5.1 Estimation of Gait Events 143
7.5.2 Estimation of Gait Parameters 144
7.5.2.1 Estimation of Orientation 144
7.5.2.2 Estimating Angles Using Accelerometers 146
7.5.2.3 Estimating Angles Using Gyroscopes 147
7.5.2.4 Fusing Accelerometer and Gyroscope Data 148
7.5.2.5 Quaternion Based Estimation of Orientation 148
7.5.2.6 Step Length Estimation 150
7.6 Conclusions 152
References 152
8 Popular Health Monitoring Systems 157
8.1 Introduction 157
8.2 Technology for Data Acquisition 157
8.3 Physiological Health Monitoring Technologies 158
8.3.1 Predicting Patient Deterioration 158
8.3.2 Ambient Assisted Living: Monitoring Daily Living Activities 163
8.3.3 Monitoring Chronic Obstructive Pulmonary Disease Patients 164
8.3.4 Movement Tracking and Fall Detection/Prevention 165
8.3.5 Monitoring Patients with Dementia 166
8.3.6 Monitoring Patients with Parkinson's Disease 168
8.3.7 Odour Sensitivity Measurement 172
8.4 Conclusions 174
References 174
9 Machine Learning for Sensor Networks 183
9.1 Introduction 183
9.2 Clustering Approaches 187
9.2.1 k-means Clustering Algorithm 187
9.2.2 Iterative Self-organising Data Analysis Technique 188
9.2.3 Gap Statistics 188
9.2.4 Density-based Clustering 189
9.2.5 Affinity-based Clustering 190
9.2.6 Deep Clustering 190
9.2.7 Semi-supervised Clustering 191
9.2.7.1 Basic Semi-supervised Techniques 191
9.2.7.2 Deep Semi-supervised Techniques 191
9.2.8 Fuzzy Clustering 192
9.3 Classification Algorithms 193
9.3.1 Decision Trees 193
9.3.2 Random Forest 194
9.3.3 Linear Discriminant Analysis 194
9.3.4 Support Vector Machines 195
9.3.5 k-nearest Neighbour 201
9.3.6 Gaussian Mixture Model 201
9.3.7 Logistic Regression 202
9.3.8 Reinforcement Learning 202
9.3.9 Artificial Neural Networks 203
9.3.9.1 Deep Neural Networks 204
9.3.9.2 Convolutional Neural Networks 205
9.3.9.3 Recent DNN Approaches 207
9.3.10 Gaussian Processes 208
9.3.11 Neural Processes 208
9.3.12 Graph Convolutional Networks 209
9.3.13 Naïve Bayes Classifier 209
9.3.14 Hidden Markov Model 210
9.3.14.1 Forward Algorithm 212
9.3.14.2 Backward Algorithm 212
9.3.14.3 HMM Design 212
9.4 Common Spatial Patterns 213
9.5 Applications of Machine Learning in BSNs and WSNs 216
9.5.1 Human Activity Detection 216
9.5.2 Scoring Sleep Stages 217
9.5.3 Fault Detection 218
9.5.4 Gas Pipeline Leakage Detection 218
9.5.5 Measuring Pollution Level 218
9.5.6 Fatigue-tracking and Classification System 218
9.5.7 Eye-blink Artefact Removal from EEG Signals 219
9.5.8 Seizure Detection 219
9.5.9 BCI Applications 219
9.6 Conclusions 219
References 220
10 Signal Processing for Sensor Networks 229
10.1 Introduction 229
10.2 Signal Processing Problems for Sensor Networks 230
10.3 Fundamental Concepts in Signal Processing 231
10.3.1 Nonlinearity of the Medium 231
10.3.2 Nonstationarity 232
10.3.3 Signal Segmentation 233
10.3.4 Signal Filtering 236
10.4 Mathematical Data Models 237
10.4.1 Linear Models 237
10.4.1.1 Prediction Method 237
10.4.1.2 Prony's Method 238
10.4.1.3 Singular Spectrum Analysis 240
10.4.2 Nonlinear Modelling 242
10.4.3 Gaussian Mixture Model 243
10.5 Transform Domain Signal Analysis 245
10.6 Time-frequency Domain Transforms 245
10.6.1 Short-time Fourier Transform 245
10.6.2 Wavelet Transform 246
10.6.2.1 Continuous Wavelet Transform 246
10.6.2.2 Examples of Continuous Wavelets 247
10.6.2.3 Discrete Time Wavelet Transform 247
10.6.3 Multiresolution Analysis 248
10.6.4 Synchro-squeezing Wavelet Transform 249
10.7 Adaptive Filtering 250
10.8 Cooperative Adaptive Filtering 251
10.8.1 Diffusion Adaptation 252
10.9 Multichannel Signal Processing 254
10.9.1 Instantaneous and Convolutive BSS Problems 255
10.9.2 Array Processing 257
10.10 Signal Processing Platforms for BANs 258
10.11 Conclusions 259
References 260
11 Communication Systems for Body Area Networks 267
11.1 Introduction 267
11.2 Short-range Communication Systems 271
11.2.1 Bluetooth 271
11.2.2 Wi-Fi 272
11.2.3 ZigBee 272
11.2.4 Radio Frequency Identification Devices 273
11.2.5 Ultrawideband 273
11.2.6 Other Short-range Communication Methods 274
11.2.7 RF Modules Available in Market 275
11.3 Limitations, Interferences, Noise, and Artefacts 275
11.4 Channel Modelling 276
11.4.1 BAN Propagation Scenarios 276
11.4.1.1 On-body Channel 276
11.4.1.2 In-body Channel 277
11.4.1.3 Off-body Channel 277
11.4.1.4 Body-to-body (or Interference) Channel 278
11.4.2 Recent Approaches to BAN Channel Modelling 278
11.4.3 Propagation Models 279
11.4.4 Standards and Guidelines 283
11.5 BAN-WSN Communications 284
11.6 Routing in WBAN 285
11.6.1 Posture-based Routing 285
11.6.2 Temperature-based Routing 286
11.6.3 Cross-layer Routing 287
11.6.4 Cluster-based Routing 288
11.6.5 QoS-based Routing 289
11.7 BAN-building Network Integration 290
11.8 Cooperative BANs 290
11.9 BAN Security 291
11.10 Conclusions 292
References 292
12 Energy Harvesting Enabled Body Sensor Networks 301
12.1 Introduction 301
12.2 Energy Conservation 302
12.3 Network Capacity 302
12.4 Energy Harvesting 303
12.5 Challenges in Energy Harvesting 304
12.6 Types of Energy Harvesting 307
12.6.1 Harvesting Energy from Kinetic Sources 308
12.6.2 Energy Sources from Radiant Sources 312
12.6.3 Energy Harvesting from Thermal Sources 312
12.6.4 Energy Harvesting from Biochemical and Chemical Sources 313
12.7 Topology Control 315
12.8 Typical Energy Harvesters for BSNs 317
12.9 Predicting Availability of Energy 318
12.10 Reliability of Energy Storage 319
12.11 Conclusions 320
References 321
13 Quality of Service, Security, and Privacy for Wearable Sensor Data 325
13.1 Introduction 325
13.2 Threats to a BAN 326
13.2.1 Denial-of-service 326
13.2.2 Man-in-the-middle Attack 327
13.2.3 Phishing and Spear Phishing Attacks 327
13.2.4 Drive-by Attack 327
13.2.5 Password Attack 328
13.2.6 SQL Injection Attack 328
13.2.7 Cross-site Scripting Attack 328
13.2.8 Eavesdropping 328
13.2.9 Birthday Attack 329
13.2.10 Malware Attack 329
13.3 Data Security and Most Common Encryption Methods 330
13.3.1 Data Encryption Standard (DES) 331
13.3.2 Triple DES 331
13.3.3 Rivest-Shamir-Adleman (RSA) 331
13.3.4 Advanced Encryption Standard (AES) 332
13.3.5 Twofish 334
13.4 Quality of Service (QoS) 334
13.4.1 Quantification of QoS 335
13.4.1.1 Data Quality Metrics 335
13.4.1.2 Network Quality Related Metrics 335
13.5 System Security 337
13.6 Privacy 339
13.7 Conclusions 339
References 340
14 Existing Projects and Platforms 345
14.1 Introduction 345
14.2 Existing Wearable Devices 347
14.3 BAN Programming Framework 348
14.4 Commercial Sensor Node Hardware Platforms 348
14.4.1 Mica2/MicaZ Motes 348
14.4.2 TelosB Mote 349
14.4.3 Indriya-Zigbee Based Platform 350
14.4.4 IRIS 350
14.4.5 iSense Core Wireless Module 351
14.4.6 Preon32 Wireless Module 351
14.4.7 Wasp Mote 352
14.4.8 WiSense Mote 352
14.4.9 panStamp NRG Mote 354
14.4.10 Jennic JN5139 354
14.5 BAN Software Platforms 355
14.5.1 Titan 355
14.5.2 CodeBlue 355
14.5.3 RehabSPOT 356
14.5.4 SPINE and SPINE2 356
14.5.5 C-SPINE 356
14.5.6 MAPS 356
14.5.7 DexterNet 356
14.6 Popular BAN Application Domains 356
14.7 Conclusions 359
References 359
15 Conclusions and Suggestions for Future Research 363
15.1 Summary 363
15.2 Future Directions in BSN Research 363
15.2.1 Smart Sensors: Intelligent, Biocompatible, and Wearable 364
15.2.2 Big Data Problem 366
15.2.3 Data Processing and Machine Learning 366
15.2.4 Decentralised and Cooperative Networks 367
15.2.5 Personalised Medicine Through Personalised Technology 367
15.2.6 Fitting BSN to 4G and 5G Communication Systems 367
15.2.7 Emerging Assistive Technology Applications 368
15.2.8 Solving Problems with Energy Harvesting 368
15.2.9 Virtual World 368
15.3 Conclusions 369
References 369
Index 373
1
Introduction
1.1 History of Wearable Technology
Earlier in history, it would take hundreds of years between breakthroughs such as eyeglasses being developed in 1286 and the abacus ring being manufactured in 1600. Today, new wearable tech innovations happen on a monthly basis, if not weekly. In the last 10 years, we have had the Google Glass, Fitbit, Oculus Rift, and countless others.
The Nuremberg egg manufactured in 1510 by Peter Henlein was one of the early portable mechanical timekeeping devices (like a watch) which had a chain to hang over the neck. An air-conditioned top hat was a wearable designed by a Victorian in the nineteenth century. In 1890, a lighting company in New York used to send girls with wearable lights onto the performance stage and to light up houses during ceremonies. In the1960s, the wearers of roulette shoes, created by Edward Thorp and Claude Shannon, used to observe the rotations of the roulette ball, tap the shoe accordingly, and then receive a vibration telling them which number to bet on. In 1963, a small portable TV screen was worn as a glass. The aviator Alberto Santos-Dumont pioneered the use of the wristwatch in 1904 as it allowed him to have his hands free while flying. This also led people to start using wristwatches. Calculator watches came onto market in 1975 and the first low-cost Walkman stereo was offered by Sony in 1979. In the 1990s, interest in the Internet of Things (IoT) started to rise. In December 1994, Steve Mann, a Canadian researcher, developed the wearable wireless Webcam. Despite its bulk, it paved the way for future IoT technologies. This required advances in artificial intelligence, which started to flourish in the 2000s.
The Sony Walkman was a clear commercial success. The Walkman and subsequent Sony Discman helped the company become an entertainment powerhouse. Over 400 million Walkman portable music players have been sold with about 200 million of those being cassette players.
However, not all products launched with a fanfare are destined for success. The commercial potential of many wearable technologies introduced in recent years are not always predictable or even achieved.
Fitbit filed for a $100 million initial public offering, but it now has to compete against a plethora of other fitness trackers on the market. The Apple Watch was been launched amidst a great deal of publicity, but it comes with no guarantees for Apple - a company that needs a lot of new revenues on a product to move the needle. Finally, the creation of the Oculus Rift virtual reality (VR) headsets could finally bring VR to the masses. The company has already been bought by Facebook for over $2 billion. Garmin, as a global positioning system (GPS), and Samsung Galaxy Gear, as a smart watch, are other popular wearables.
What is clear is that, based on the history of wearable technology, devices that move the masses are far and between. The successes that do make it, however, can change the world and generate chart-topping returns. Meanwhile, people's needs change over time, and include entertainment, activity, sport, and now most importantly health. This brings wearables such as Quell to the market. When strapped on the body Quell predicts and detects the onset of chronic pain and stimulates nerves to block pain signals to the brain. Other wearables to measure blood alcohol content, athletic performance, blood sugar, heart rate, and many other bioindicators rapidly came to the market as the desire for health monitoring grew. This may become more demanding as the interest in personal medicine grows.
1.2 Introduction to BSN Technology
Wearable technology including sensors, sensor networks, and the associated devices has opened its space in a variety of applications. Long-term, noninvasive, and nonintrusive monitoring of the human body through collecting as much biometric data and state indicators as possible is the major goal of healthcare wearable technology developers. Patients suffering from diabetes need a simple noninvasive tool to monitor their blood sugar on an hourly basis. Those suffering from seizure require the necessary instrumentation to alarm them before any seizure onset to prevent them from fall injury. The stroke patients need their heart rate recorded constantly. These are only a small number of examples which show how crucial and necessary wearable healthcare systems can be.
At the Wearable Technology Conference in 2018, the winners of seven wearable device producers were introduced. These winners include the best ones in Lifestyle with the objective of 'play stress away'; Sports and Fitness for making a football performance device, healthcare for developing a smart eyewear with assistive artificial intelligence capabilities for the blind and visually impaired; Industrial for designing a unique smart and connected industry 4.0 safety shoe; Smart Clothing Challenge for the nonintrusive acquisition of heart signals that will enable pervasive health monitoring, emotional state assessment, drowsiness detection, and identity recognition; Smart Lamp, which allows you to move the light in any direction without moving the lamp; and Connected Living Challenge, for creating accessories linking braintech with fashion design. Headpieces and earrings use electroencephalography (EEG) technology, capturing and providing users with brain data, allowing them to be conscious of their mental state in real time, for example for reducing anxiety and depression or increasing focus or relaxation of the user [1]. This simple example together with the above examples clearly show the diversity in applications of wearable technology. The aim of this book is therefore to familiarise readers with sensors, connections, signal processing tools and algorithms, electronics, communication systems, and networking protocols as well as many applications of wearable devices for the monitoring of mental, metabolic, physical, and physiological states of the human body.
Disease prevention, patient monitoring, and disable and elderly homecare have become the major objectives for investment in social health and public wellbeing. According to the World Health Organization (WHO), an ageing population is becoming a significant problem and degenerative brain diseases, such as dementia and depression, are increasingly seen in people while a bad lifestyle is causing millions of people to suffer from obesity or chronic diseases. It is thus reasonable to expect that this circumstance will only contribute to an ongoing decline in the quality of services (QoSs) provided by an already overloaded healthcare system [2]. A remote low-cost monitoring strategy, therefore, would significantly promote social and clinical wellbeing. This can only be achieved if sufficient reliably recorded information from the human body is available. Such information may be metabolic, biological, physiological, behavioural, psychological, functional, or motion-related.
On the other hand, the development of mobile telephone systems since the early 1990s and its improvement till now together with the availability of large size archiving and wideband communication channels significantly increase the chance of achieving the above objectives without hospitalising the caretakers in hospitals and care units for a long time. This may be considered a revolution in human welfare. More effective and efficient data collection from the human body has therefore a tremendous impact and influence on healthcare and the technology involved. The state of a patient during rest, walking, working, and sleeping can be well recognised if all the biomarkers of the physiological, biological, and behavioural changes of human body can be measured and processed. This requirement sparks the need for deployment of a multisensor and multimodal data collection system on the body. A body sensor network (BSN) therefore is central to a complete solution for patient monitoring and healthcare. Several key applications benefit from the advanced integration of BSNs, often called body area networks (BANs), with the new mobile communication technology [3, 4].
The main applications of BSNs are expected to appear in the healthcare domain, especially for the continuous monitoring and logging of vital parameters of elderly people or patients suffering from degenerative diseases such as dementia or chronic diseases such as diabetes, asthma, and heart attacks. As an example, a BAN network on a patient can alert the hospital, even before they have a heart attack, through measuring changes in their vital signs, or placing it on a diabetic patient could auto-inject insulin through a pump as soon as their insulin level declines.
The IEEE 802.15 Task Group 6 (BAN) is developing a communication standard optimised for reliable low-power devices and operation on, in, or around the human body (but not limited to humans) to serve a variety of applications including medical, consumer electronics/personal entertainment, and security [5]. This was approved on 22 July 2011 and the first meeting of IEEE 802.15 wireless personal area network (WPAN) was held on 3 March 2017.
The BSN technology benefits from developments in various areas of sensors, automation, communications, and more closely the vast advances in wired and wireless sensor networks (WSNs) for short- and long-range communications and industrial control. For...
