Vision, Sensing and Analytics: Integrative Approaches (Intelligent Systems Reference Library, 207)

✍ Scribed by Md Atiqur Rahman Ahad (editor), Atsushi Inoue (editor)

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 416
• Category: Library

✦ Synopsis

This book serves as the first guideline of the integrative approach, optimal for our new and young generations. Recent technology advancements in computer vision, IoT sensors, and analytics open the door to highly impactful innovations and applications as a result of effective and efficient integration of those. Such integration has brought to scientists and engineers a new approach ―the integrative approach. This offers far more rapid development and scalable architecting when comparing to the traditional hardcore developmental approach.

Featuring biomedical and healthcare challenges including COVID-19, we present a collection of carefully selective cases with significant added- values as a result of integrations, e.g., sensing with AI, analytics with different data sources, and comprehensive monitoring with many different sensors, while sustaining its readability.

✦ Table of Contents

Foreword
Preface
Contents
1 Deep Architectures in Visual Transfer Learning
1.1 Introduction
1.2 Transfer Learning
1.3 Datasets
1.3.1 ImageNet
1.3.2 Caltech-256 Dataset
1.4 Deep Architectures
1.4.1 VGG16
1.4.2 VGG19
1.4.3 Xception
1.4.4 DenseNet121
1.4.5 ResNet50
1.5 Experiments and Results
1.5.1 VGG16
1.5.2 VGG19
1.5.3 Xception
1.5.4 DenseNet121
1.5.5 ResNet50
1.5.6 All at Once
1.6 Conclusion and Future Work
References
2 Deep Reinforcement Learning: A New Frontier in Computer Vision Research
2.1 Introduction
2.2 Reinforcement Learning
2.2.1 Basics
2.2.2 Solving MDPs Using Reinforcement Learning Algorithms
2.3 Deep Reinforcement Learning
2.4 Deep Reinforcement Learning Methods
2.4.1 Attention Aware Deep Reinforcement Learning
2.4.2 Deep Progressive Reinforcement Learning
2.4.3 Multi-agent Deep Reinforcement Learning
2.5 Deep Reinforcement Learning Applications in Computer Vision
2.5.1 Visual Search and Tracking
2.5.2 Video Captioning and Summarization
2.5.3 Image Processing and Understanding
2.5.4 Action Detection, Recognition and Prediction
2.5.5 Robotics
2.6 Research Challenges
2.6.1 Exploration Vs. Exploitation
2.6.2 Multi Agent Systems
2.6.3 Real World Conditions
2.6.4 Transfer Learning
2.7 Conclusion and Future Research Direction
2.7.1 Deep Inverse Reinforcement Learning
2.7.2 Robotic Vision: Visual Grasping and Navigation
2.7.3 Few Shot and Zero Shot Learning
References
3 Deep Learning for Data-Driven Predictive Maintenance
3.1 Introduction
3.2 Maintenance
3.2.1 Predictive Maintenance
3.3 Sensors Commonly Used for Predicative Maintenance
3.4 Sensor Data Analysis and Machine Learning
3.4.1 Logistic Regression
3.4.2 k-Nearest Neighbors
3.4.3 Artificial Neural Network
3.4.4 Support Vector Machine
3.5 Limitations of Machine Learning Algorithms Used for Predictive Maintenance
3.6 Deep Learning Methods
3.6.1 Deep Artificial Neural Network
3.6.2 Deep Convolutional Neural Network
3.6.3 Deep Recurrent Neural Network
3.6.4 Deep Auto-encoders
3.6.5 Deep Belief Network
3.7 Advantages of Deep Learning in Predictive Maintenance
3.8 Deep Learning for Predictive Maintenance
3.9 Conclusion and Future Perspectives of Deep Learning Based Data-Driven Predictive Maintenance
3.9.1 Conclusion
3.9.2 Future Perspectives of Deep Learning Based Data-Driven Predictive Maintenance
3.10 Glossary
References
4 Multi-criteria Fuzzy Goal Programming Under Multi Uncertainty
4.1 Introduction
4.2 Preliminary Preparation
4.2.1 Fuzzy Number
4.2.2 Membership Function
4.2.3 GP
4.2.4 Fuzzy GP
4.2.5 Notation
4.3 Fuzzy Multi-criteria Linear Programming
4.3.1 Purposes of this Research
4.3.2 Definitions of Fuzzy Sets
4.3.3 Building the Goal Model by Fuzzy Random Regression
4.4 Formulate Constraints Equations
4.4.1 Definition of Constraints
4.4.2 Solution
4.5 Application
4.5.1 Case Introduction
4.5.2 Solution
4.5.3 Construction of Goal Equations
4.5.4 Formulation of Constraints' Equations
4.5.5 Solution of the Multi-criteria Linear Goal Models
4.6 Conclusions
References
5 Skeleton-Based Human Action Recognition on Large-Scale Datasets
5.1 Introduction
5.2 Large-Scale Skeletal Datasets
5.2.1 NTU-60 Dataset
5.2.2 NTU-120 Dataset
5.2.3 Kinetics-400 Dataset
5.2.4 Other Challenging Datasets
5.3 Deep Learning-Based Methods for SHAR
5.3.1 Co-occurrence Feature Learning-Based Methods for SHAR
5.3.2 LSTM-Based Methods for SHAR
5.3.3 Multi-Task Learning (MTLN)-Based Approaches for SHAR
5.3.4 Graph Convolutional Network Based Approaches for SHAR
5.3.5 Adversarial Approaches
5.3.6 Hidden Markov Model (HMM)
5.3.7 Joint-Centered Local Information (JOLO)
5.4 Performance Analysis and Discussion
5.5 Conclusion and Future Direction
References
6 Sensor-Based Human Activity and Behavior Computing
6.1 Introduction
6.2 Statistical Learning-Based Approaches
6.2.1 Feature Extraction
6.2.2 Feature Selection Technique
6.2.3 Statistical Classifiers and Factors
6.3 Data Collection for HAR
6.4 Sensor-Based Benchmark Datasets in HAR
6.4.1 Database Repositories
6.4.2 Sensor-Based Datasets for Healthcare
6.4.3 Sensor-Based Datasets for Physical Fitness and Sports Activities
6.4.4 Sensor-Based Datasets for Household Activities
6.4.5 Sensor-Based Datasets for Device Usage Activities
6.4.6 Sensor-Based Datasets for Transportation Activities
6.4.7 Sensor-Based Datasets for Fall Detection
6.4.8 Smart Wearable Sensing-Based Datasets
6.4.9 Smart Device-Based Datasets (Mobilephone/Tablets)
6.5 Overview of Classification Problems and Performance Measures in HAR
6.6 Deep Learning-Based Approaches
6.7 Conclusion and Future Direction: Challenges to Overcome
References
7 Radar-Based Non-Contact Physiological Sensing
7.1 Introduction
7.2 Basics of Radar
7.2.1 Radar Principle of Operation and Waveforms
7.2.2 Continuous-Wave (CW) Radar
7.2.3 Frequency Modulated Continuous Wave (FMCW) Radar
7.2.4 Pulse Doppler Radar
7.3 Traditional Methods of Respiratory Monitoring
7.3.1 Need for Non-contact Remote Respiration Monitoring and Potential Application Areas
7.3.2 Patients with Burn Injury
7.3.3 Sleep Monitoring
7.3.4 Unobtrusive Identity Authentication
7.3.5 Elderly Monitoring
7.3.6 Occupancy Sensing
7.3.7 Unmanned Aerial Vehicle (UAV)-Based Life Sensing Radar
7.4 Basics of Doppler Radar for Non-contact Cardiopulmonary Monitoring
7.4.1 Theory of CW Doppler Radar for Cardiopulmonary Monitoring
7.4.2 Receiver Architecture
7.4.3 Single Channel and Quadrature Receiver
7.4.4 Data Acquisition and Signal Processing
7.4.5 FMCW Signal Processing for Vital Signs Sensing
7.5 Future Challenges
7.5.1 Multiple Subjects Respiration Monitoring
7.5.2 Random Body Movement Cancellation
7.5.3 Identity Authentication of Multiple Subject
7.5.4 Breathing Diversity in Radar Authentication
7.5.5 Radar Authentication Data Security
7.6 Radar Physiological Sensing for New Researchers
7.6.1 Step 1: Radar Data Channel Imbalance Minimization
7.6.2 Step 2: Calculating Imbalance Values
7.6.3 Step 3: Channel Combining Method
7.6.4 Step 4: Separation of Cardiac and Respiratory Component
7.6.5 Step 5: Analyze Waveform for Breathing and Heart Rate Extraction
7.7 Conclusion
References
8 Biomedical Radar and Antenna Systems for Contactless Human Activity Analysis
8.1 Introduction
8.2 Contactless Monitoring for Healthcare
8.3 Radar-Based Sensing
8.3.1 Data Processing
8.3.2 Data Classification
8.4 Radar-Based Human Activity Analysis
8.4.1 Posture Recognition
8.4.2 Search and Rescue Purposes
8.4.3 Sleep Monitoring
8.4.4 Human Activity Recognition
8.4.5 Identification of Individuals
8.4.6 Monitoring of Vital Signs
8.4.7 Occupancy Monitoring
8.4.8 Fall Detection
8.5 Future Challenges
8.5.1 Development of Suitable Approaches
8.5.2 Development of Readily Available Data Sets
8.5.3 Well-Designed Clinical Trials
8.5.4 Compensation of Noise and Artifacts
8.5.5 Elimination of Clutter
8.5.6 Multi-person Monitoring
8.5.7 Generalization of Motion
8.5.8 Subject-Specific Classification
8.5.9 Recognition of Complex Human Activities
8.5.10 Search and Rescue Operation
8.6 Conclusion
References
9 Contactless Monitoring for Healthcare Applications
9.1 Introduction
9.2 Cardiac and Blood Related Monitoring
9.2.1 Contactless ECG
9.2.2 Remote Photoplethysmography (rPPG)
9.2.3 Heart Rate and Heart Rate Variability Measurement
9.2.4 Blood Pressure Monitoring
9.2.5 Blood Flow Monitoring
9.2.6 Oxygen Saturation (SpO2) Monitoring
9.3 Respiratory Monitoring
9.3.1 Respiration Rate Estimation
9.3.2 Sleep Monitoring from Breathing Pattern
9.4 Neurological Monitoring
9.4.1 Symptoms Detection in Neuro-Degenerative Diseases
9.4.2 Home Monitoring for Alzheimer's Patients
9.4.3 Rehabilitation of Post-stroke Patients
9.5 Blood Glucose Monitoring
9.6 Discussion
9.7 Conclusion
References
10 Personalized Patient Safety Management: Sensors and Real-Time Data Analysis
10.1 Introduction
10.1.1 Overview of a Personalized Health Care System
10.1.2 Current Trends on the Development of the Personalized Medical Device
10.1.3 Sensor-Based Medical Device
10.1.4 Sensor-Based Personalized Healthcare Devices, Safety, and Well-Being
10.2 Development of Sensors-Based Personalized Medical Devices
10.2.1 Designing of Sensor-Based Medical Devices
10.2.2 System Architecture and Application
10.2.3 Monitoring the Device Activity
10.2.4 Data Processing and Feature Extraction
10.2.5 Process of Data Analysis
10.3 Implementation of Personalized Devices in the Physiological Systems
10.3.1 Challenges During Implementation Procedures
10.3.2 The Working Mechanism of the Device
10.3.3 Monitoring System
10.4 Data Security for the Patient
10.4.1 Health Records in Electronic Forms and Health Information Systems
10.4.2 Security and Privacy
10.4.3 Secure Transmission
10.5 Regulatory Aspect for Commercialization
10.5.1 Legislative Policies in the Sensor-Based Personalized Healthcare System
10.5.2 Government Agencies Shaping Personalized Medicine
10.5.3 Habitats for Medicare and Medicaid Services (CMS)
10.5.4 Food and Drug Administration (FDA)
10.5.5 Public Institutes of Health (NIH)
10.5.6 European Medicines Agency (EMA)
10.6 Conclusions and Future Perspectives
References
11 Electrical Impedance Tomography Based Lung Disease Monitoring
11.1 Introduction
11.1.1 Applications of EIT
11.1.2 Benefits of EIT
11.1.3 Fundamental Characteristics of EIT
11.1.4 Drawbacks of EIT
11.2 Lung Disease Monitoring
11.2.1 Acute Respiratory Distress Syndrome (ARDS)
11.2.2 Chronic Obstructive Pulmonary Disease (COPD)
11.2.3 Cystic Fibrosis (CF)
11.2.4 Pneumonia
11.2.5 Pleural Effusion
11.3 Discussion and Future Challenges
11.3.1 Development of Readily Available Data Sets
11.3.2 Well-Designed Clinical Trials
11.3.3 Studying the Effect of Diverse Testing Conditions
11.3.4 Compensation of Noise and Errors
11.3.5 3D Image Reconstruction
11.3.6 Verification of Phantom-Based Findings
11.3.7 Other Possibilities
11.4 Conclusion
References
12 Image Analysis with Machine Learning Algorithms to Assist Breast Cancer Treatment
12.1 Introduction
12.2 Related Work
12.2.1 Methods to Treat Breast Cancer
12.2.2 Mammography
12.2.3 Computer-Assisted Technology in Breast Cancer Treatment
12.3 Proposed Image Processing Mechanism
12.3.1 Detecting Malignant Cells
12.3.2 Guiding Surgical Procedures
12.4 Evaluation
12.4.1 Assumptions
12.4.2 Tools and Languages Used
12.4.3 Training and Testing
12.4.4 Mammogram Images Used
12.4.5 Experimental Results
12.4.6 Guiding Surgical Procedures
12.5 Conclusion and Future Scopes
Appendix A
References
13 Role-Framework of Artificial Intelligence in Combating the COVID-19 Pandemic
13.1 Introduction
13.2 Data
13.3 Role Framework of Artificial Intelligence
13.3.1 Early Trace-Out, Detection, and Diagnosis
13.3.2 Disease Surveillance, Control, Awareness Build-Up, and Prevention
13.3.3 Monitoring the Treatment and Predicting the Risk of Developing a Severe Case
13.3.4 Screening and Helping Patients Through Chatbots
13.3.5 Service Management Through Intelligence Drones and Robots
13.3.6 Management of Stress and the Spread of Rumors Through Social Network
13.3.7 Understanding the Virus Through Analysis of Protein–Protein Interactions
13.3.8 Speeding up the Vaccine and Drug Discoveries and Development
13.3.9 Continuation of Education and Prediction of Economic Loss
13.4 Commercial Applications of AI
13.5 Conclusion
References
14 Time Series Analysis for CoVID-19 Projection in Bangladesh
14.1 Introduction
14.2 Materials and Mathematical Models to Epidemic
14.2.1 Materials
14.2.2 Logistic Growth Model
14.2.3 SIS (Susceptible-Infectious-Susceptible) Epidemic Model
14.2.4 SIR Epidemic Model
14.2.5 SEIR (Susceptible-Exposed-Infectious-Recovered) Epidemic Model
14.3 Analysis of CoVID-19 Cases
14.3.1 Analysis of Confirmed Cases, Death Cases and Recovered Cases in Bangladesh
14.3.2 Analysis of Daily New Cases and Cumulative Confirmed Cases
14.4 Projection of CoVID-19 in Bangladesh
14.4.1 Application of Logistic Growth Model
14.4.2 Application of SIR Model
14.5 Conclusion
References
15 Challenges Ahead in Healthcare Applications for Vision and Sensors
15.1 Introduction
15.2 Application and Challenges of Computer Vision in Healthcare
15.2.1 Facial Expression Detection
15.2.2 Sign Language Translation
15.2.3 Detection of Safety Critical Incidents
15.2.4 Visual Prosthesis
15.2.5 Surgical Assistance Technology
15.2.6 Rehabilitation Aids
15.2.7 Medical Imaging
15.3 Application and Challenges of Sensor Technology in Healthcare
15.3.1 Wearable Sensor
15.3.2 Epidermal Sensor
15.3.3 Contact Lens Biosensor
15.3.4 Implantable Sensor
15.4 Conclusion
References