Ambulation Analysis in Wearable ECG (eBook)

Ambulation Analysis in Wearable ECG 2
Preface 6
Acknowledgments 9
Contents 10
1 Introduction 13
1.1 Basics of Electrocardiogram (ECG) 13
1.2 Artifacts in ECG 17
1.2.1 EMG Noise 17
1.2.2 Baseline Wandering 17
1.2.3 Powerline Interference 18
1.2.4 Motion Artifacts 18
1.3 Ambulatory Monitoring 18
1.4 Challenges in Ambulatory ECG Processing 20
1.5 Mathematical Model of Ambulatory ECG Signal 21
1.6 Tour of the Book 22
2 Review of ECG Analysis 26
2.1 QRS Detection Methods 27
2.2 Delineation of Wave Boundaries 29
2.3 Beat Alignment 31
2.4 Noise Reduction in ECG 32
2.5 Detection of Body Posture Changes 34
2.6 Overview of Wearable ECG Recorders 35
2.7 Analysis of Ambulation in ECG 36
3 Hardware Development of Wearable ECGDevices* 38
3.1 Introduction 38
3.2 Basics of Personal ECG Instruments 39
3.2.1 System Modules and Operation 39
3.2.2 System Requirements 40
3.3 Electrodes 40
3.4 Signal Conditioning 41
3.4.1 Implementation Using General-Purpose ICs 42
3.4.2 ASIC (Application-Speci c Integrated Circuit) Design forSignal Conditioning 44
3.5 Analog to Digital Converter 50
3.6 Digital Modules 53
3.6.1 Microcontroller 53
3.6.2 Data Storage 53
3.6.3 Data Retrieval 54
3.7 Discussion 54
4 Calibration of Locket 55
4.1 Calibration Requirements 56
4.2 Experimental Set-up 57
4.3 Calibration Technique 58
4.4 Results and Discussion 59
5 Data Acquisition 63
5.1 Introduction 63
5.2 Commonplace Body Movement Activities 64
5.3 Activity Transition 65
5.4 Motion Sensing 67
5.4.1 Data Collection using Accelerometer 67
5.4.2 Processing of Accelerometer Data 69
5.5 Variation of Activity Levels 70
5.6 Protocols for Treadmill Tests 70
6 Detection of Activity Transition 72
6.1 Introduction 73
6.2 Transition Detection 75
6.3 Experimental Results 79
6.4 Discussion 86
7 Activity Recognition 87
7.1 Introduction 88
7.2 Nonparametric Classi cation 90
7.2.1 Preprocessing 92
7.2.2 Principal Component Analysis (PCA) 94
7.2.3 Supervised Learning of Body Movement 96
7.2.4 Activity Classi cation 98
7.2.5 Removal of Motion Artifacts 99
7.3 Parametric classi cation 99
7.3.1 Pre-processing 101
7.3.2 Feature Extraction 103
7.3.3 Hidden Markov Model (HMM) and Training 105
7.3.4 Activity Classi cation 107
7.4 Experimental Results 108
7.4.1 PCA-based Recognition 108
7.4.2 HMM-based Recognition 119
7.5 Discussion 128
8 Impact of Ambulation 130
8.1 Introduction 131
8.2 Derivation of Impact Signal 132
8.3 Synchronization of Impact and Motion Data 133
8.4 Experimentations 134
8.4.1 Experiments on the Treadmill 134
8.4.2 Experiments with Motion Sensors 137
8.5 Discussions 145
9 Conclusions 149
9.1 Conclusions 149
9.2 Scopes for Future Work 151
References 154
Index 163