Modeling and Optimization of Signals Using Machine Learning Techniques
Description
Explore the power of machine learning to revolutionize signal processing and optimization with cutting-edge techniques and practical insights in this outstanding new volume from Scrivener Publishing.
Modeling and Optimization of Signals using Machine Learning Techniques is designed for researchers from academia, industries, and R&D organizations worldwide who are passionate about advancing machine learning methods, signal processing theory, data mining, artificial intelligence, and optimization. This book addresses the role of machine learning in transforming vast signal databases from sensor networks, internet services, and communication systems into actionable decision systems. It explores the development of computational solutions and novel models to handle complex real-world signals such as speech, music, biomedical data, and multimedia.
Through comprehensive coverage of cutting-edge techniques, this book equips readers with the tools to automate signal processing and analysis, ultimately enhancing the retrieval of valuable information from extensive data storage systems. By providing both theoretical insights and practical guidance, the book serves as a comprehensive resource for researchers, engineers, and practitioners aiming to harness the power of machine learning in signal processing.
Whether for the veteran engineer, scientist in the lab, student, or faculty, this groundbreaking new volume is a valuable resource for researchers and other industry professionals interested in the intersection of technology and agriculture.
More details
Other editions
Additional editions
Persons
Chandra Singh is an assistant professor in the Department of Electronics and Communication Engineering at Sahyadri College of Engineering and Management, Mangalore, India, and is pursuing a PhD from VTU Belagavi, India. He has four patents, he has published over 25 papers in scientific journals, and he is the editor of seven books.
Rathishchandra R. Gatti, PhD, is an associate professor at Jawaharlal Nehru University, Delhi, India. With over 20 years of industrial, research, and teaching experience under his belt, he also has four patents, has published over 40 papers in scientific journals, and is the editor of seven research books and one journal.
K.V.S.S.S.S.SAIRAM, PhD, is a professor and Head of the Electronics and Communication Engineering Department at the NMAM Institute of Technology, Nitte, India. He has 25 years of experience in teaching and research and has published over 50 papers in international journals and conferences. He is also a reviewer for several journals.
Manjunatha Badiger, PhD, is an assistant professor at Sahyadri College of Engineering and Management, Adyar, Mangalore, Karnataka, India. He has over 12 years of experience in academics, research, and administration. He earned his PhD in machine learning in 2024 at Visvesvaraya Technological University.
Naveen Kumar S., MTech, is an assistant professor at the Sahyadri College of Engineering and Management. Previously he was an assistant professor at JSS Academy of Technical Education, Noida, India. He obtained his MTech in automotive electronics from Sri Jayachamarajendra College of Engineering, Mysore, India.
Varun Saxena, PhD, received his PhD in electromagnetic ion traps from IIT Delhi, New Delhi, in 2018. He is currently an assistant professor at the School of Engineering, Jawaharlal Nehru University, New Delhi.
Content
Preface xix
1 Land Use and Land Cover Mapping of Remotely Sensed Data Using Fuzzy Set Theory-Related Algorithm 1
Adithya Kumar and Shivakumar B.R.
1.1 Introduction 2
1.2 Image Classification 5
1.3 Unsupervised Classification 7
1.4 Supervised Classification 8
1.5 Overview of Fuzzy Sets 9
1.6 Methodology 11
1.7 Results and Discussion 16
1.8 Conclusion 21
2 Role of AI in Mortality Prediction in Intensive Care Unit Patients 23
Prabhudutta Ray, Sachin Sharma, Raj Rawal and Dharmesh Shah
2.1 Introduction 24
2.2 Background 24
2.3 Objectives 25
2.4 Machine Learning and Mortality Prediction 26
2.5 Discussions 34
2.6 Conclusion 34
2.7 Future Work 35
2.8 Acknowledgments 35
2.9 Funding 35
2.10 Competing Interest 35
3 A Survey on Malware Detection Using Machine Learning 41
Devika S. P., Pooja M. R. and Arpitha M. S.
3.1 Background 41
3.2 Introduction 42
3.3 Literature Survey 44
3.4 Discussion 53
3.5 Conclusion 53
4 EEG Data Analysis for IQ Test Using Machine Learning Approaches: A Survey 55
Bhoomika Patel H. C., Ravikumar V. and Pavan Kumar S. P.
4.1 Related Work 57
4.2 Equations 62
4.3 Classification 64
4.4 Data Set 65
4.5 Information Obtained by EEG Signals 69
4.6 Discussion 70
4.7 Conclusion 72
5 Machine Learning Methods in Radio Frequency and Microwave Domain 75
Shanthi P. and Adish K.
5.1 Introduction 76
5.2 Background on Machine Learning 77
5.3 ML in RF Circuit Modeling and Synthesis 86
5.4 Conclusion 93
6 A Survey: Emotion Detection Using Facial Reorganization Using Convolutional Neural Network (CNN) and Viola-Jones Algorithm 97
Vaibhav C. Gandhi, Dwij Kishor Siyal, Shivam Pankajkumar Patel and Arya Vipesh Shah
6.1 Introduction 98
6.2 Review of Literature 99
6.3 Report on Present Investigation 101
6.4 Algorithms 102
6.5 Viola-Jones Algorithm 104
6.6 Diagram 105
6.7 Results and Discussion 107
6.8 Limitations and Future Scope 111
6.9 Summary and Conclusion 111
7 Power Quality Events Classification Using Digital Signal Processing and Machine Learning Techniques 115
E. Fantin Irudaya Raj and M. Balaji
7.1 Introduction 116
7.2 Methodology for the Identification of PQ Events 117
7.3 Power Quality Problems Arising in the Modern Power System 118
7.4 Digital Signal Processing-Based Feature Extraction of PQ Events 124
7.5 Feature Selection and Optimization 129
7.6 Machine Learning-Based Classification of PQ Disturbances 131
7.7 Summary and Conclusion 141
8 Hybridization of Artificial Neural Network with Spotted Hyena Optimization (SHO) Algorithm for Heart Disease Detection 145
Shwetha N., Gangadhar N., Mahesh B. Neelagar, Sangeetha N. and Virupaxi Dalal
8.1 Introduction 146
8.2 Literature Survey 147
8.3 Proposed Methodology 149
8.4 Artificial Neural Network 152
8.5 Software Implementation Requirements 163
8.6 Conclusion 170
9 The Role of Artificial Intelligence, Machine Learning, and Deep Learning to Combat the Socio-Economic Impact of the Global COVID-19 Pandemic 173
Biswa Ranjan Senapati, Sipra Swain and Pabitra Mohan Khilar
9.1 Introduction 174
9.2 Discussions on the Coronavirus 175
9.3 Bad Impacts of the Coronavirus 180
9.4 Benefits Due to the Impact of COVID-19 186
9.5 Role of Technology to Combat the Global Pandemic COVID-19 190
9.6 The Role of Artificial Intelligence, Machine Learning, and Deep Learning in COVID-19 198
9.7 Related Studies 203
9.8 Conclusion 203
10 A Review on Smart Bin Management Systems 209
Bhoomika Patel H. C., Soundarya B. C. and Pooja M. R.
10.1 Introduction 209
10.1.1 Internet of Things (IoT) 210
10.2 Related Work 211
10.3 Challenges, Solution, and Issues 213
10.4 Advantages 216
11 Unlocking Machine Learning: 10 Innovative Avenues to Grasp Complex Concepts 219
K. Vidhyalakshmi and S. Thanga Ramya
11.1 Regression 220
11.2 Classification 222
11.3 Clustering 227
11.4 Clustering (k-means) 227
11.5 Reduction of Dimensionality 230
11.6 The Ensemble Method 233
11.7 Transfer of Learning 240
11.8 Learning Through Reinforcement 241
11.9 Processing of Natural Languages 242
11.10 Word Embeddings 242
11.11 Conclusion 243
12 Recognition Attendance System Ensuring COVID-19 Security 245
Praveen Kumar M., Ramya Poojary, Saksha S. Bhandary and Sushmitha M. Kulal
12.1 Introduction 246
12.2 Literature Survey 246
12.3 Software Requirements 248
12.4 Hardware Requirements 249
12.5 Methodology 252
12.6 Building the Database 253
12.7 Pi Camera for Extracting Face Features 255
12.8 Real-Time Testing on Raspberry Pi 256
12.9 Contactless Body Temperature Monitoring 256
12.10 Raspberry-Pi Setting Up an SMTP Email 258
12.11 Uploading to the Database 259
12.12 Updating the Website 260
12.13 Report Generation 260
12.14 Result 262
12.15 Discussion 267
12.16 Conclusion 267
13 Real-Time Industrial Noise Cancellation for the Extraction of Human Voice 271
Vinayprasad M. S., Chandrashekar Murthy B. N. and Yashwanth S. D.
13.1 Introduction 272
13.2 Literature Survey 273
13.3 Methodology 275
13.4 Experimental Results 278
13.5 Conclusion 280
14 Machine Learning-Based Water Monitoring System Using IoT 283
T. Kesavan, E. Kaliappan, K. Nagendran and M. Murugesan
14.1 Introduction 283
14.2 Smart Water Monitoring System 284
14.3 Sensors and Hardware 286
14.4 PowerBI Reports 288
14.5 Conclusion 291
15 Design and Modelling of an Automated Driving Inspector Powered by Arduino and Raspberry Pi 295
Raghunandan K. R., Dilip Kumar K., Krishnaraj Rao N.S. Krishnaprasad Rao and Bhavya K.
15.1 Introduction 296
15.2 Literature Survey 296
15.3 Results 306
15.4 Conclusion 309
16 Kalman Filter-Based Seizure Prediction Using Concatenated Serial-Parallel Block Technique 313
Purnima P. S. and Suresh M.
16.1 Introduction 314
16.2 Prior Work 314
16.3 Proposed Method 316
16.4 Serial-Parallel Block Concatenation Approach 318
16.5 Algorithm 319
16.6 Kalman Filter 320
16.7 Results and Discussion 321
16.8 Conclusion 323
17 Current Advancements in Steganography: A Review 327
Mallika Garg, Jagpal Singh Ubhi and Ashwani Kumar Aggarwal
17.1 Introduction 328
17.2 Evaluation Parameters 329
17.3 Types of Steganography 330
17.4 Traditional Steganographic Techniques 332
17.5 CNN-Based Steganographic Techniques 336
17.6 GAN-Based Steganographic Techniques 338
17.7 Steganalysis 340
17.8 Applications 341
17.9 Dataset Used for Steganography 341
17.10 Conclusion 344
18 Human Emotion Recognition Intelligence System Using Machine Learning 349
Bhakthi P. Alva, Krishma Bopanna N., Prajwal S., Varun A. Naik and Lahari Vaidya
18.1 Introduction 350
18.2 Literature Review 350
18.3 Problem Statement 352
18.4 Methodology 353
18.5 Results 355
18.6 Applications 355
18.7 Conclusion 357
18.8 Future Work 357
19 Computing in Cognitive Science Using Ensemble Learning 361
Om Prakash Singh
19.1 Introduction 362
19.2 Recognition of Human Activities 363
19.3 Methodology 366
19.4 Applying the Boosting-Based Ensemble Learning 369
19.5 Human Activity Features Computability 373
19.6 Conclusion 378
References 378
About the Editors 383
Index 385
1
Land Use and Land Cover Mapping of Remotely Sensed Data Using Fuzzy Set Theory-Related Algorithm
Adithya Kumar* and Shivakumar B.R.
NMAM Institute of Technology, NITTE, Udupi, Karnataka, India
Abstract
All bodies, planets, living beings, and inanimate objects emit electromagnetic radiation, and the amount and type of radiation emitted depend largely on their temperatures. Electromagnetic radiation may be emitted by an object or may come from another body and could be reflected by it. There are operational satellite systems that sample virtually every region of the electromagnetic spectrum, with spatial resolutions from 0.5 to 5,000 m. The scientific community's interest in spatiotemporal studies of global change, environmental monitoring, and human impacts involves the use of remote sensing data. Remote sensing systems, particularly those located on satellites, provide a repetitive and synoptic vision of Earth, which is of great interest in monitoring and analyzing human activities and their impacts. The activities include the evaluation and monitoring of the environment like urban growth and hazardous waste; detection and monitoring of global changes, deforestation, global warming, and exploration of non-renewable resources and their land use; civil engineering;, the acquisition of satellite imagery from relevant sources; the preprocessing of satellite imagery as per requirement; and the development and classification of satellite data using fuzzy C-means (FCM), modified FCM, K-means, and fuzzy inference system (FIS) techniques.
Keywords: FCM, modified FCM, K-means, FIS techniques
1.1 Introduction
In this chapter, an introduction to remote sensing theory, the differences between practical and ideal remote sensing systems and the Landsat 8 data characteristics and concepts of electromagnetic radiation are presented [1, 2]. Remote sensing can be broadly defined as the science of obtaining useful information about physical objects and/or the environment through recording and measuring energy patterns from the sensor systems. The information about distant objects and the environment is obtained through the electromagnetic radiations emitted or reflected by the Earth's surface features, which are of interest. Multiple data users that make use of the ideal remote sensing system would be able to acquire the required data at a higher speed with less or no expense of any required area of interest [3, 4]. Given this information, multiple data users would be able to make decisions on managing and observing the Earth's surface features. A real remote sensing system encounters a lot of problems in each stage of the remote sensing system. The following section explains the limitations of all ideal remote sensing systems. The spectral distribution of reflected and/or emitted electromagnetic energy is not uniform, and it varies from time and place on different Earth features. Passive remote sensing systems entirely depend on the energy from the sun. Solar energy is non-uniform with respect to time and place. The sources used in the remote sensing systems are also non- uniform with respect to wavelength and time. Therefore, there is always a need to calibrate these sources depending on the mission, time, and location. In a real remote sensing system, there is always a need for atmospheric and radiometric corrections due to atmospheric errors. The atmosphere always modifies the energy emitted and reflected by the Earth's surface features. Eliminating the atmospheric effects is important in those applications involving observations of the same geographical area. Different materials do not reflect and emit energy uniquely. There is spectral overlapping in the recorded data. The spectral response plays an important role in detecting, identifying, and analyzing the Earth's surface materials. A good understanding of energy interactions is required to obtain information about the required features. From the above points, it is clear that super sensors do not exist. The sensors cannot be sensitive to all possible wavelengths and have limits with respect to sensitivity. The sensors record the objects according to their spatial resolution. The ability of the sensors to differentiate the smallest object and separate it from its surroundings is called spatial resolution. The spatial resolution depends on the heterogeneity of the ground area being sensed. The digital images obtained consist of mixed pixels causing problems in identifying the land cover classes. Depending on the application, the choice of sensors varies from airborne vehicles to space stations. Photographic systems produce images of higher spatial resolution, but spectral sensitivity is absent in those sensors. Therefore, to obtain information on all the land cover classes, sensors with higher spectral resolution are required rather than sensors with higher spatial resolution. However, sensors having higher spectral resolution are expensive. Proper image processing techniques are required to visualize if the sensor data are of low resolution. Therefore, there are always tradeoffs between cost and resolution. Remote sensing systems generally are of two types: passive and active. An active remote sensing system possesses an illumination system or source of energy. It directs this energy to the objects to measure the energy. Some examples of active sensor systems are radars and laser scanners. The passive remote sensing system utilizes the sun's energy to record the emitted and/or reflected values of the surface features. Landsat series sensors are an example for passive remote sensing systems. The sensors that generate artificial radiation to the Earth's surface features are called active sensors. Radio detection and ranging or radar is used to detect the Earth's surface features using the pulses of electromagnetic radiation in microwave ranges. These radiations are not affected by weather conditions such as clouds, fog, and wind. The reflected signal intensity is used to extract the information. These types of sensors using self-generated radiation are useful for specific satellite missions and operations [5, 6].
Sensors that use natural light and the thermal radiation of the Earth's features to extract the information are called passive. This information is greatly affected by the seasons, time of capture, and height of the satellite. The Landsat satellite is equipped with spectrometers that measure the signals at a spectrum of bands. Therefore, multispectral channels are obtained, which provide greater information. Satellite imagery consisting of pixel values are the measures of the intensity of the electromagnetic radiation of the Earth's surface features. The values of brightness differ in the image based on time, type of area, and sensors. Certain properties of imagery like resolution should be known. The individual elements of the digital image are called pixels. The size of the pixel and the type of sensor determine the spatial resolution. The length of the edges of the pixels is used to measure resolution. The number of spectral channels is defined as the spectral resolution of the image. The reflection of Earth surface features occurs in different wavelengths. The human eyes can recognize only the visible spectrum of radiations. The land area, which has high reflections in particular wavelengths, is sensed by the satellite sensors. Most of the Earth observation satellites have between three and eight bands. The linear imaging self-scanning sensor III (LISS-III) has four spectral bands. The higher spectral resolution allows larger features to be extracted.
The time interval the satellite takes to reach the same area is called temporal resolution. The temporal resolution is determined by the orbit of the satellite, the altitude, and the sensor's characteristics. Generally, Earth observation satellites have a repetition rate of 14-16 days. Landsat 7 has 16 days. The highest temporal resolution is for meteorological satellites, which is 15 minutes [7, 8].
The surface reflectance characteristics of three main Earth surface features are discussed here:
- Vegetation: The chlorophyll pigment in leaves absorbs radiation in red and blue wavelengths but reflects in the green wavelength. The reflectance characteristics vary with wavelength [9, 10, 16, 18].
- Water: The water absorbs most of the radiation in the longer wavelengths. The water reflects in shorter wavelengths and hence blue in appearance. In the near-infrared wavelengths, the water is darker in color. The reflectance varies due to the depth of the water, the materials in the water, and the type of water [12].
- Soil: The nature of the soil determines the reflection and emission. The presence of moisture decreases the reflectance. The reflectance of soil is present in a large range of wavelengths depending upon the soil [13].
1.1.1 Overview on Landsat 8
Landsat 8 is an American Earth observation satellite launched in 2013. It is the seventh satellite to reach orbit successfully. The operational land imager (OLI) sensor consists of nine capturing devices, and thermal infrared sensor (TIS) has two capturing devices. The thermal bands measure the surface temperature and lighter pixels represent hot temperatures. Thermal bands provide a lot of information regarding heat units present in urban regions. The Landsat 8 orbit height is 705 km above the Earth's surface. It collects 400 scenes of 185-km wide regions in one day, and an archive is made...
