Machine Learning in Signal Processing

Applications, Challenges, and the Road Ahead

Sudeep Tanwar, Anand Nayyar, Rudra Rameshwar, Sudeep Tanwar, Anand Nayyar, Rudra Rameshwar

• Publisher: Chapman and Hall/CRC
• Year: 2021
• Print ISBN: 9780367618902
• eBook ISBN: 9781000487817
• Topic: Technology & Engineering
• Subtopic: Computer Engineering
• Index: Technology & Engineering

Machine Learning in Signal Processing: Applications, Challenges, and the Road Ahead offers a comprehensive approach toward research orientation for familiarizing signal processing (SP) concepts to machine learning (ML).

ML, as the driving force of the wave of artificial intelligence (AI), provides powerful solutions to many real-world technical and scientific challenges. This book will present the most recent and exciting advances in signal processing for ML.

The focus is on understanding the contributions of signal processing and ML, and its aim to solve some of the biggest challenges in AI and ML.

FEATURES

• Focuses on addressing the missing connection between signal processing and ML
• Provides a one-stop guide reference for readers
• Oriented toward material and flow with regards to general introduction and technical aspects
• Comprehensively elaborates on the material with examples and diagrams

This book is a complete resource designed exclusively for advanced undergraduate students, post-graduate students, research scholars, faculties, and academicians of computer science and engineering, computer science and applications, and electronics and telecommunication engineering.

1 Introduction to Signal Processing and Machine Learning

Kavitha Somaraj

Higher Colleges of Technology

DOI: 10.1201/9781003107026-1

CONTENTS

1. 1.1 Introduction
2. 1.2 Basic Terminologies
   1. 1.2.1 Signal Processing
      1. 1.2.1.1 Continuous and Discrete Signals
      2. 1.2.1.2 Sampling and Quantization
      3. 1.2.1.3 Change of Basis
      4. 1.2.1.4 Importance of Time Domain and Frequency Domain Analyses
   2. 1.2.2 Machine Learning
3. 1.3 Distance-Based Signal Classification, Nearest Neighbor Classifier, and Hilbert Space
   1. 1.3.1 Distance-Based Signal Classification
      1. 1.3.1.1 Metric Space
      2. 1.3.1.2 Normed Linear Space
      3. 1.3.1.3 Inner Product Space
   2. 1.3.2 Nearest Neighbor Classification
   3. 1.3.3 Hilbert Space
4. 1.4 Fusion of Machine Learning in Signal Processing
5. 1.5 Benefits of Adopting Machine Learning in Signal Processing
6. 1.6 Conclusion
7. References

1.1 Introduction

Signal processing has gained a lot of appreciation in the modern era with its applications growing virtually in all walks of life such as communications, entertainment, control, and environment, just to mention a few. Digital signal processing (DSP) is a main driver of a digital revolution that introduced compact disks, DVDs, digital cameras, mobile phones, and countless technological devices. The clinical diagnosis of various health ailments would be less efficient if medical equipment such as X-rays, electroencephalograph, and electrocardiogram analyzers that depend on DSP did not exist. Recent advancements have led to hi-tech gadgets and wearables based on human machine interface (HMI) technology such as Google Glasses, Xbox games, and Fitbits to name a few.

Signal processing can be defined as the process of extracting useful information from a signal. It depends on the nature of the signal and the type of information that needs to be extracted. It deals with the transformation, analysis, and synthesis of signals or information. The purpose of signal processing is to modify the given signal such that the quality of information is improved in some well-defined meaning [1]. Signal processing is classified as analog signal processing and digital signal processing. An analog signal varies continuously with time, and as described in [2], the term analog appears to have stemmed from the analog computers used prior to 1980. Those computers solved linear differential equations by means of connecting physical (electronic) differentiators and integrators using old-style telephone operator patch cords. However, present-day computers and technologies rely on digital signals, thereby making analog signals obsolete. DSP deals with the analysis of discrete or digitized sampled signals [3], and it has a number of advantages when compared to analog signal processing such as high processing accuracy, stability, and reliability.

DSP evolved in 1960, and since then, the use of DSP has been growing continuously, because of the development of powerful and efficient methods, particularly filter design techniques and fast Fourier transform (FFT) algorithms, opening several application areas [4].The traditional digital signal processing technique utilized specialized hardware to process signals which included data acquisition, signal transformation, analysis, synthesis, filtering, evaluation and identification, etc. in order to extract information [5]. However, with the advancement of technology and living standards, DSP has evolved even further, and the most recent development is the fusion of signal processing and machine learning referred to as SPML. The present-day wearable gadgets such as Google Glasses etc. mentioned earlier depend on human machine interaction and brain computer interface signals that must translate signals from the user’s brain into messages or commands which can be achieved through SPML.

The current SMAC (Social, Mobile, Analytics, Cloud) technology trend paves the way to a future in which intelligent machines, networked processes, and big data are brought together [6]. A large volume of data or information is generated through various sources in this virtual world, and therefore, the modern society is facing the challenge of “information overload.” There is an immediate need to deal with sheer volumes of data, which poses a significant constraint to traditional DSP techniques. Moreover, the data that needs to be processed is available in different formats. These challenges have led to the adoption of powerful machine learning algorithms. As described in [7], machine learning adequately fits the constraints and solution requirements posed by DSP problems: from computational efficiency, online adaptation, and learning with limited supervision to their ability to combine heterogeneous information, to incorporate prior knowledge about the problem, or to interact with the user to achieve improved performance. Few academic researchers have adopted ML and related algorithms for various applications such as detecting building or construction defects using image processing [8], automatically generating highlights of a broadcasted cricket match [9], secure data analytics [10], block-chain-based smart applications [11], traffic management [12], and distributed big data analytics [13]. Moreover, SPML is currently used in various real-world applications that depend on statistical data preprocessing and feature extraction techniques such as image processing, object detection, biometrics, and voice recognition, to mention a few. For these reasons, it is crucial for engineers, scientists, technologists, and alike to have the right knowledge in order to...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Preface
• Editors
• Contributors
• 1. Introduction to Signal Processing and Machine Learning
• 2. Learning Theory (Supervised/Unsupervised) for Signal Processing
• 3. Supervised and Unsupervised Learning Theory for Signal Processing
• 4. Applications of Signal Processing
• 5. Dive in Deep Learning: Computer Vision, Natural Language Processing, and Signal Processing
• 6. Brain–Computer Interfacing
• 7. Adaptive Filters and Neural Net
• 8. Adaptive Decision Feedback Equalizer Based on Wavelet Neural Network
• 9. Intelligent Video Surveillance Systems Using Deep Learning Methods
• 10. Stationary Signal, Autocorrelation, and Linear and Discriminant Analysis
• 11. Intelligent System for Fault Detection in Rotating Electromechanical Machines
• 12. Wavelet Transformation and Machine Learning Techniques for Digital Signal Analysis in IoT Systems
• Index