Machine learning espousal in signal processing : applications, challenges and road ahead

Cover
Half Title
Title Page
Copyright Page
Table of Contents
Preface
Editors
Contributors
1. Introduction to Signal Processing and Machine Learning
	1.1 Introduction
	1.2 Basic Terminologies
		1.2.1 Signal Processing
			1.2.1.1 Continuous and Discrete Signals
			1.2.1.2 Sampling and Quantization
			1.2.1.3 Change of Basis
			1.2.1.4 Importance of Time Domain and Frequency Domain Analyses
		1.2.2 Machine Learning
	1.3 Distance-Based Signal Classification, Nearest Neighbor Classifier, and Hilbert Space
		1.3.1 Distance-Based Signal Classification
			1.3.1.1 Metric Space
			1.3.1.2 Normed Linear Space
			1.3.1.3 Inner Product Space
		1.3.2 Nearest Neighbor Classification
		1.3.3 Hilbert Space
	1.4 Fusion of Machine Learning in Signal Processing
	1.5 Benefits of Adopting Machine Learning in Signal Processing
	1.6 Conclusion
	References
2. Learning Theory (Supervised/Unsupervised) for Signal Processing
	2.1 Introduction
		2.1.1 Signal Processing
	2.2 Machine Learning
		2.2.1 Why Do We Need ML for Signal Processing?
		2.2.2 Speaker ID – A Utilization of ML Calculations in Sign Handling
		2.2.3 Discourse and Audio Processing
		2.2.4 Discourse Recognition
		2.2.5 Listening Devices
		2.2.6 Independent Driving
		2.2.7 Picture Processing and Analysis
		2.2.8 Wearables
		2.2.9 Information Science
		2.2.10 Wireless Systems and Networks
	2.3 Machine Learning Algorithms
	2.4 Supervised Learning
	2.5 Unsupervised Learning
	2.6 Semi-Supervised Learning
	2.7 Reinforcement Learning
	2.8 Use Case of Signal Processing Using Supervised and Unsupervised Learning
		2.8.1 Features and Classifiers
		2.8.2 Linear Classifiers
		2.8.3 Decision Hyperplanes
		2.8.4 Least Squares Methods
		2.8.5 Mean Square Estimation
		2.8.6 Support Vector machines
		2.8.7 Non-Linear Regression
		2.8.8 Non-Linearity of Activation Functions
			2.8.8.1 Sigmoid Function
			2.8.8.2 Rectified Linear Unit (ReLU)
		2.8.9 Classification
			2.8.9.1 Linear Classification
			2.8.9.2 Two-Class Classification
			2.8.9.3 Geometrical Interpretation of Derivatives
			2.8.9.4 Multiclass Classification: Loss Function
		2.8.10 Mean Squared Error
		2.8.11 Multilabel Classification
		2.8.12 Gradient Descent
			2.8.12.1 Learning Rate
		2.8.13 Hyperparameter Tuning
			2.8.13.1 Validation
		2.8.14 Regularization
			2.8.14.1 How Does Regularization Work?
		2.8.15 Regularization Techniques
			2.8.15.1 Ridge Regression (L2 Regularization)
		2.8.16 Lasso Regression (L1 Regularization)
		2.8.17 K-Means Clustering
		2.8.18 The KNN Algorithm
		2.8.19 Clustering
		2.8.20 Clustering Methods
	2.9 Deep Learning for Signal Data
		2.9.1 Traditional Time Series Analysis
		2.9.2 Recurrence Domain Analysis
		2.9.3 Long Short-Term Memory Models for Human Activity Recognition
		2.9.4 External Device HAR
		2.9.5 Signal Processing on GPUs
		2.9.6 Signal Processing on FPGAs
		2.9.7 Signal Processing is coming to the Forefront of Data Analysis
	2.10 Conclusion
	References
3. Supervised and Unsupervised Learning Theory for Signal Processing
	3.1 Introduction
		3.1.1 Supervised Learning
		3.1.2 Unsupervised Learning
		3.1.3 Reinforcement Learning
		3.1.4 Semi-Supervised Learning
	3.2 Supervised Learning Method
		3.2.1 Classicfiation Problems
		3.2.2 Regression Problems
		3.2.3 Examples of Supervised Learning
	3.3 Unsupervised Learning Method
		3.3.1 Illustrations of Unsupervised Learning
	3.4 Semi-Supervised Learning Method
	3.5 Binary Classification
		3.5.1 Different Classes
		3.5.2 Classification in Preparation
			3.5.2.1 Logistic Regression Model
			3.5.2.2 Odds Ratio
			3.5.2.3 Logit Function
			3.5.2.4 The Sigmoid Function
			3.5.2.5 Support Vector Machines
			3.5.2.6 Maximum Margin Lines
	3.6 Conclusion
	References
4. Applications of Signal Processing
	4.1 Introduction
	4.2 Audio Signal Processing
		4.2.1 Machine Learning in Audio Signal Processing
			4.2.1.1 Spectrum and Cepstrum
			4.2.1.2 Mel Frequency Cepstral Coefficients
			4.2.1.3 Gammatone Frequency Cepstral Coefficients
			4.2.1.4 Building the Classifier
	4.3 Audio Compression
		4.3.1 Modeling and Coding
		4.3.2 Lossless Compression
		4.3.3 Lossy Compression
		4.3.4 Compressed Audio with Machine Learning Applications
	4.4 Digital Image Processing
		4.4.1 Fields Overlapping with Image Processing
		4.4.2 Digital Image Processing System
		4.4.3 Machine Learning with Digital Image Processing
			4.4.3.1 Image Classification
			4.4.3.2 Data Labelling
			4.4.3.3 Location Detection
	4.5 Video Compression
		4.5.1 Video Compression Model
		4.5.2 Machine Learning in Video Compression
			4.5.2.1 Development Savings
			4.5.2.2 Improving Encoder Density
	4.6 Digital Communications
		4.6.1 Machine Learning in Digital Communications
			4.6.1.1 Communication Networks
			4.6.1.2 Wireless Communication
			4.6.1.3 Smart Infrastructure and IoT
			4.6.1.4 Security and Privacy
			4.6.1.5 Multimedia Communication
		4.6.2 Healthcare
			4.6.2.1 Personalized Medical Treatment
			4.6.2.2 Clinical Research and Trial
			4.6.2.3 Diagnosis of Disease
			4.6.2.4 Smart Health Records
			4.6.2.5 Medical Imaging
			4.6.2.6 Drug Discovery
			4.6.2.7 Outbreak Prediction
		4.6.3 Seismology
			4.6.3.1 Interpreting Seismic Observations
			4.6.3.2 Machine Learning in Seismology
		4.6.4 Speech Recognition
		4.6.5 Computer Vision
		4.6.6 Economic Forecasting
	4.7 Conclusion
	References
5. Dive in Deep Learning: Computer Vision, Natural Language Processing, and Signal Processing
	5.1 Deep Learning: Introduction
	5.2 Past, Present, and Future of Deep- Learning
	5.3 Natural Language Processing
		5.3.1 Word Embeddings
			5.3.1.1 Word2vec
		5.3.2 Global Vectors for Word Representation
		5.3.3 Convolutional Neural Networks
		5.3.4 Feature Selection and Preprocessing
			5.3.4.1 Tokenization
			5.3.4.2 Stop Word Removal
			5.3.4.3 Stemming
			5.3.4.4 Lemmatization
		5.3.5 Named Entity Recognition
	5.4 Image Processing
		5.4.1 Introduction to Image Processing and Computer Vision
			5.4.1.1 Scene Understanding
		5.4.2 Localization
		5.4.3 Smart Cities and Surveillance
		5.4.4 Medical Imaging
		5.4.5 Object Representation
		5.4.6 Object Detection
	5.5 Audio Processing and Deep Learning
		5.5.1 Audio Data Handling Using Python
		5.5.2 Spectrogram
		5.5.3 Wavelet- Based Feature Extraction
		5.5.4 Current Methods
			5.5.4.1 Audio Classification
			5.5.4.2 Audio Fingerprinting
			5.5.4.3 Feature Extraction
			5.5.4.4 Speech Classification
			5.5.4.5 Music Processing
			5.5.4.6 Natural Sound Processing
			5.5.4.7 Technological Tools
	5.6 Conclusion
	References
6. Brain–Computer Interfacing
	6.1 Introduction to BCI and Its Components
		6.1.1 BCI Components
	6.2 Framework/Architecture of BCI
	6.3 Functions of BCI
		6.3.1 Correspondence and Control
		6.3.2 Client State Checking
	6.4 Applications of BCI
		6.4.1 Healthcare
			6.4.1.1 Prevention
			6.4.1.2 Detection and Diagnosis
			6.4.1.3 Rehabilitation and Restoration
		6.4.2 Neuroergonomics and Smart Environment
		6.4.3 Neuromarketing and Advertisement
		6.4.4 Pedagogical and Self-Regulating Oneself
		6.4.5 Games and Entertainment
		6.4.6 Security and Authentication
	6.5 Signal Acquisition
		6.5.1 Invasive Techniques
			6.5.1.1 Intracortical
			6.5.1.2 ECoG and Cortical Surface
		6.5.2 Noninvasive Techniques
			6.5.2.1 Magneto-encephalography (MEG)
			6.5.2.2 fMRI (functional Magnetic Resonance Imaging)
			6.5.2.3 fNIRS (functional Near-Infrared Spectroscopy)
			6.5.2.4 EEG (Electroencephalogram)
	6.6 Electrical Signal of BCI
		6.6.1 Evoked Potential (EP) or Evoked Response
		6.6.2 Event-Related Desynchronization and Synchronization
	6.7 Challenges of BCI and Proposed Solutions
		6.7.1 Challenges of Usability
		6.7.2 Technical Issues
		6.7.3 Proposed Solutions
			6.7.3.1 Noise Removal
			6.7.3.2 Disconnectedness of Multiple Classes
	6.8 Conclusion
	References
7. Adaptive Filters and Neural Net
	7.1 Introduction
		7.1.1 Adaptive Filtering Problem
	7.2 Linear Adaptive Filter Implementation
		7.2.1 Stochastic Gradient Approach
		7.2.2 Least Square Estimation
	7.3 Nonlinear Adaptive Filters
		7.3.1 Volterra-Based Nonlinear Adaptive Filter
	7.4 Applications of Adaptive Filter
		7.4.1 Biomedical Applications
			7.4.1.1 ECG Power-Line Interference Removal
			7.4.1.2 Maternal-Fetal ECG Separation
		7.4.2 Speech Processing
			7.4.2.1 Noise Cancelation
		7.4.3 Communication Systems
			7.4.3.1 Channel Equalization in Data Transmission Systems
			7.4.3.2 Multiple Access Interference Mitigation in CDMA
		7.4.4 Adaptive Feedback Cancellation in Hearing Aids
	7.5 Neural Network
		7.5.1 Learning Techniques in ANN
	7.6 Single and Multilayer Neural Net
		7.6.1 Single-Layer Neural Networks
		7.6.2 Multilayer Neural Net
	7.7 Applications of Neural Networks
		7.7.1 ECG Classicafition
			7.7.1.1 Methodology
		7.7.2 Speech Recognition
			7.7.2.1 Methodology
		7.7.3 Communication Systems
			7.7.3.1 Mobile Station Location Identification Using ANN
			7.7.3.2 ANN-Based Call Handoff Management Scheme for Mobile Cellular Network
			7.7.3.3 A Hybrid Path Loss Prediction Model based on Artificial  Neural Networks
			7.7.3.4 Classification of Primary Radio Signals
			7.7.3.5 Channel Capacity Estimation Using ANN
	7.8 Conclusion
	References
8. Adaptive Decision Feedback Equalizer Based on Wavelet Neural Network
	8.1 Introduction
	8.2 System Model
		8.2.1 Channel Equalization
		8.2.2 Decision Feedback Equalization
	8.3 Wavelet Neural Network
		8.3.1 Wavelet Analysis
		8.3.2 Wavelet Neural Network
	8.4 Multidimensional Wavelet Neural Network
	8.5 Proposed WNN DFE Architecture
		8.5.1 Equalizer Architecture
		8.5.2 Cuckoo Search Optimization
		8.5.3 CSO-Based Training of WNN DFE
		8.5.4 Simulation Results and Discussion
			8.5.4.1 MSE Performance
			8.5.4.2 Effect of EVR
			8.5.4.3 Effect of Time-Varying Channel
			8.5.4.4 BER Performance Evaluation
	8.6 Conclusion
	References
9. Intelligent Video Surveillance Systems Using Deep Learning Methods
	9.1 Introduction
		9.1.1 Deep Learning
		9.1.2 Deep Learning – Past, Present, and Future
		9.1.3 Recent Methodologies
		9.1.4 Concepts Used in Deep Learning
			9.1.4.1 Convolutional Neural Networks (CNN)
	9.2 Natural Language Processing Using Deep Learning
		9.2.1 Introduction to Natural Language Processing (NLP)
		9.2.2 Word-Vector Representations (Simple Word, Multiword Prototypes, and Global Contexts)
			9.2.2.1 Word Vector Representation
			9.2.2.2 Simple Word2VectorRepresentation
			9.2.2.3 Learning Representation through Backpropagation
			9.2.2.4 Natural Language Tasks for Text Classification
			9.2.2.5 Natural Language Tasks for Image Description Generation
	9.3 Machine Translation Using Gated Recurrent Neural Networks (GRNN) and Long Short-Term Memory (LSTM)
		9.3.1 Gated Recurrent Units (GRUs)
		9.3.2 Long Short-Term Memory (LSTM)
		9.3.3 Results Analysis
	9.4 Image Processing Using Deep Learning Algorithms
		9.4.1 Introduction to Image Processing and Computer Vision
		9.4.2 Data Preparation for Image Processing Tasks
		9.4.3 Classification Algorithms with Applications
	9.5 Lightweight Deep Convolution Neural Network Architecture (LW-DCNN)
		9.5.1 Introduction
		9.5.2 Architecture
		9.5.3 Results
			9.5.3.1 Comparison Analysis
	9.6 Improved Unified Model for Moving Object Detection
		9.6.1 Introduction
		9.6.2 Object Detection Architecture
		9.6.3 Results
		9.6.4 Comparison Analysis
		9.6.5 Applications to Human Action Recognition
	9.7 Wavelet-Based Feature Extraction Methods and Application to Audio Signals
		9.7.1 Introduction to Discrete Wavelet Transform Techniques
		9.7.2 Wavelet-Based Feature Selection Methods
		9.7.3 Hybrid Feature Extraction Method for Classification
		9.7.4 Results
		9.7.5 Various Applications of Audio Signals
	9.8 Conclusion
	References
10. Stationary Signal, Autocorrelation, and Linear and Discriminant Analysis
	10.1 Introduction
	10.2 Fundamentals of Linear Algebra and Probability Theory
		10.2.1 What is Linear Algebra?
			10.2.1.1 Important Concepts in Linear Algebra for Machine Learning
			10.2.1.2 Role of Linear Algebra in Machine Learning
		10.2.2 Probability Theory
			10.2.2.1 What Is Probability?
			10.2.2.2 The Mathematics of Probability
			10.2.2.3 Independence and Conditional Independence
	10.3 Basic Concepts of Machine Learning
	10.4 Supervised and Unsupervised ML Techniques for Digital Signal Processing
		10.4.1 What Is Signal Processing?
		10.4.2 Machine Learning (ML) Concepts
	10.5 Applications of Signal Based Identification Using Machine Learning Approach
		10.5.1 ML for Audio Classification
		10.5.2 Audio Signals Classification
		10.5.3 ML for Image Processing
	10.6 Applications of ML Methods in Optical Communications
	10.7 Conclusion
	References
11. Intelligent System for Fault Detection in Rotating Electromechanical Machines
	11.1 Introduction
	11.2 Related Works
	11.3 Asynchronous Machines
	11.4 Electromechanical Faults
		11.4.1 Bearing Fault
		11.4.2 Broken Rotor Bar Fault
		11.4.3 Eccentricity Fault
		11.4.4 Misalignment Fault
	11.5 Methods for Detecting Anomalies
		11.5.1 Definition
		11.5.2 Importance of Anomaly Detection
		11.5.3 Some Techniques for Anomaly Detection
	11.6 Frequency Signatures
	11.7 The MCSA Measurement Method
		11.7.1 Modeling of the Stator Current of the Asynchronous Machine
	11.8 Variants of the ESPRIT Method
	11.9 MOS (Order Selection Model)
		11.9.1 Principle
		11.9.2 Mathematical Expressions
		11.9.3 Results Obtained by Each of the MOS Algorithms
			11.9.3.1 Conclusion
	11.10 Intelligent Defect Classification Algorithms
		11.10.1 Artificial Neuronal Networks and Genetic Algorithms (ANN-AG)
			11.10.1.1 Artificial Neural Networks
			11.10.1.2 Genetic Algorithms (GA)
		11.10.2 Fusion ANN et AG
		11.10.3 Association of Two Architectures
		11.10.4 Support Vectors Machine (SVM)
			11.10.4.1 How It Works
		11.10.5 K-Nearest Neighbors (K-NN)
		11.10.6 Extreme Learning Machines
			11.10.6.1 Principle or Algorithm
	11.11 Simulation and Analysis of Results
		11.11.1 High-Resolution Estimation Methods
			11.11.1.1 Preparation of Simulation Data
			11.11.1.2 Frequency Error Analysis
			11.11.1.3 Amplitude Error Analysis
			11.11.1.4 Interpretations on Frequency Analysis
			11.11.1.5 Interpretations on Amplitude Analysis
			11.11.1.6 Interpretations on Frequency and Amplitude Analysis
			11.11.1.7 Interpretation of Algorithm Execution Times
			11.11.1.8 Conclusion
		11.11.2 Fault Classification Algorithms
			11.11.2.1 Artificial Neural Networks and Genetic Algorithms
			11.11.2.2 Conclusion
		11.11.3 Vector Machine Supports
			11.11.3.1 Simulation in the Time Domain
			11.11.3.2 Simulation in the Frequency Domain
		11.11.4 K-Nearest Neighbors
			11.11.4.1 Simulation in the Time Domain
			11.11.4.2 Simulation in the Frequency Domain
		11.11.5 Extreme Learning Machine
			11.11.5.1 Simulation in the Time Domain
			11.11.5.2 Simulation in the Frequency Domain
		11.11.6 Comparative Table of the Different Algorithms Developed in Time and Frequency
			11.11.6.1 Comparison of Intelligent Fault Classification Algorithms in the Time and Frequency Domain
	11.12 Conclusion
	References
12. Wavelet Transformation and Machine Learning Techniques for Digital Signal Analysis in IoT Systems
	12.1 Introduction
	12.2 Digital Signal Processing Techniques for IoT Devices
		12.2.1 Fourier Transform
		12.2.2 Wavelet Transform
			12.2.2.1 Continuous Wavelet Transform (CWT)
			12.2.2.2 Discrete Wavelet Transformation (DWT)
			12.2.2.3 Computation of Discrete Wavelet Transform
	12.3 Machine Learning and Deep Learning Techniques for Time Series Analysis in IoT
		12.3.1 Time Series Classification Algorithms
		12.3.2 Time Series Classification Using Deep Learning
	12.4 Comparison for Morlet, Mexican Hat, Frequency B-Spline Wavelet Toward the Classification of ECG Signal
		12.4.1 Mexican Wavelet Transform
		12.4.2 Morlet Wavelet Transform
		12.4.3 Frequency B-Spline Wavelet Transform
	12.5 Conclusion
	References
Index