This thesis presents a series of neural network classifiers for object detection in both optical and infrared images. The focus of this work is on efficient and accurate solutions. The thesis discusses the evolution of the highly efficient and tiny network Binary Classification Vision Transformer (BC-ViT) and how through thoughtful modifications and improvements, the BC-ViT can be utilized for tasks of increasing complexity. Chapter 2 discusses the creation of BC-ViT and its initial use case for underwater image classification of optical images. The BC-ViT is able to complete its task with an accuracy of 99.29\% while being comprised of a mere 15,981 total trainable parameters. Chapter 3, Waste Multi-Class Vision Transformer (WMC-ViT), introduces the usefulness of mindful algorithm design for the realm of multi-class classification on a mutually exclusive dataset. WMC-ViT shows that the task oriented design strategy allowed for a network to achieve an accuracy score of 94.27\% on a five class problem while still maintaining a tiny parameter count of 35,492. The final chapter demonstrates that by utilizing functional blocks of BC-ViT, a simple and effective target detection algorithm for infrared images can be created. The Edge Infrared Vision Transformer (EIR-ViT) showed admirable results with a high IoU …

PM2.5 particle sensing has been done using surface acoustic wave based sensor for two different frequencies. Due to mass loading and elasticity loading on the sensor's surface, the center frequency of the sensor shifts. The particle concentration can be tracked based on that frequency shift. The fit factor test has been conducted using higher frequency SAW sensor. The consist results has been achieved for particle sensing and fit factor test with SAW based sensor.

My thesis is to understand the concept of compressive sensing algorithms. Compressive sensing will be a future alternate technique for the Nyquist rate, specific to some applications where sparsity property plays a major role. Software implementation of compressive sensing (CS) takes more time to reconstruct a signal from CS measurements, so we use the orthogonal matching pursuit and basis pursuit algorithms. We have used an image size of 256x256 is used for reconstruction and also implemented a field-programmable gate array (FPGA) of the orthogonal matching pursuit using an image.

This thesis covers applications of machine learning to the fields of remote sensing and environmental monitoring. First, a generalized background on the concepts, tools, and methods used throughout the remainder of the research project are introduced. Chapter 3 covers the implementation of artificial neural networks to improve low-cost particulate matter sensing networks using collocated high-quality sensors with varying dataset parameters. In Chapter 4, an attention-enhanced LSTM-Convolutional neural network is presented to reconstruct satellite-based aerosol optical depth data lost to atmospheric interference. Chapter 5 applies attention mechanisms and convolutional neural networks to the reconstruction and upsampling of satellite-based land surface temperature maps. Chapter 6 presents a model employing geospatial techniques and machine learning methods with a combination of ground-based and remote sensing data to produce a daily ultra-high resolution 30 meter mapping of the PM2.5 concentration across Denton County, Texas.

The syndrome source coding for lossless data compression with side information based on fixed-length linear block codes is the main emphasis of this work. We demonstrate that the source entropy rate can be achieved for syndrome source coding with side information when the sources are correlated. Next, we examine employing LDPC codes to apply the channel and syndrome concepts in order to satisfy the Slepian Wolf limit. Our findings indicate that irregular codes perform significantly better when the compression ratio is larger. Additionally, we looked at how well different applications performed when running on two different mobile networks. We have tested those applications which are used in our day-to-day life. Our main focus is to make wireless communication much easier. We know that nowadays data is increasing which led to increase in the transfer of data. There are a lot of errors while doing so like channel error, bit error rate, jitter, etc. To overcome such kind of problems compression and decompression should be done effectively without any complexity to achieve a high performance ratio.

This thesis creates a new method for the ethical splitting of data as well as improvements to neural network architectures to increase performance. Ethical dataset splitting should be based on statistics from the data, this prevents artificial manipulation of the data that helps or hurts the performance of a network. This bias introduced to the dataset can also be present by using the popular method of randomly splitting data into datasets. To remove bias from dataset splitting, the splits of a dataset must be based on statistics from the data. Improving neural network architectures to increase performance is very important for a wide range of applications, especially for classification of heartbeats. Every improvement matters, especially when the application means that any errors could put the life of a person in danger. These advancements being applied to heartbeat classification have exciting implications for saving thousands of lives and billions of dollars. The presented methods can also be expanded to a wide variety of applications and adapted to different types of data as increasing performance and splitting up datasets is important in all fields of machine learning.

This thesis presents an optimized drive signal control system for a 2.5 GHz Doherty power amplifier (PA). The designed system enables independent control of the amplitudes and phases of the drive signals fed to the inputs of two parallel PAs. This control system is demonstrated here for Doherty PA architecture with a combiner network which is used as an impedance inversion between the path of two parallel connected PAs. Independent control of the inputs is achieved by incorporating a variable attenuator (VA) and a variable phase shifter (VPS) in each of the two parallel paths. Integrating VA and VPS allows driving varying power levels with an arbitrary phase difference between the individual parallel PAs. A Combiner network consists of a quarter-wave transmission line at the output of the main power amplifier, which is used to invert the impedance between the main and peaking transistor. The specific VA (Qorvo QPC6614) and VPS (Qorvo QPC2108) components that are used for the test system provide an amplitude attenuation range from 0.5 dB to 31.5 dB with a step size of 0.5 dB and a phase range from 0◦ to 360◦ for a step size of 5.6◦at the intended operating frequency of 2.5 GHz, …

In the vast field of signal processing, machine learning is rapidly expanding its domain into all realms. As a constituent of this expansion, this thesis presents contributive work on advancements in machine learning algorithms by building on the shoulder of giants. The first chapter of this thesis contains enhancements to a CNN (convolutional neural network) for better classification of heartbeat arrhythmia. The network goes through a two stage development, the first being augmentations to the network and the second being the implementation of dropout. Chapter 2 involves the combination of CNN and LSTM (long short term memory) networks for the task of short-term energy use data regression. Exploiting the benefits of two of the most powerful neural networks, a unique, novel neural network is created to effectually predict future energy use. The final section concludes this work with directions for future works.

Modern machine learning techniques focus on extremely deep and multi-pathed networks, resulting in large memory and computational requirements. This thesis explores techniques for designing efficient convolutional networks including pixel shuffling, depthwise convolutions, and various activation fucntions. These techniques are then applied to two image processing domains: single-image super-resolution and image compression. The super-resolution model, TinyPSSR, is one-third the size of the next smallest model in literature while performing similar to or better than other larger models on representative test sets. The efficient deep image compression model is significantly smaller than any other model in literature and performs similarly in both computational cost and reconstruction quality to the JPEG standard.

The goal of the proposed device and software architecture is to apply the functionality of mixed reality (MR) in order to make a virtual environment that is more accessible to the visually-impaired. We propose a glove-based system for MR that will use finger and hand movement tracking along with tactile feedback so that the visually-impaired can interact with and obtain a more detailed sense of virtual objects and potentially even virtual environments. The software architecture makes current MR frameworks more accessible by augmenting the existing software and extensive 3D model libraries with both the interfacing of the glove-based system and the audibly navigable user interface (UI) of a virtual environment we have developed. We implemented a circuit with finger flexion/extension tracking for all 5 fingers of a single hand and variable vibration intensities for the vibromotors on all 5 fingertips of a single hand. The virtual environment can be hosted on a Windows 10 application. The virtual hand and its fingers can be moved with the system's input and the virtual fingertips touching the virtual objects trigger vibration motors (vibromotors) to vibrate while the virtual objects are being touched. A rudimentary implementation of picking up and moving virtual objects inside …

As technology has advanced, data has become more and more important. The more breakthroughs are achieved, the more data is needed to support them. As a result, more storage is required in the system's memory. Compression is therefore required. Before it can be stored, the data must be compressed. To ensure that information is not lost, efficient compression is necessary. This also makes sure that there is no redundancy in the data that is being kept and stored. Compressive sensing has emerged as a new field of compression thanks to developments in sparse optimization. Rather than relying just on compression and sensing formulations, the theory blends the two. The objective of this thesis is to analyze the concept of compressive sensing and to study several reconstruction algorithms. Additionally, a few of the algorithms were put into practice. This thesis also included a model of the ECG, which is vital in determining the health of the heart. For the most part, the ECG is utilized to diagnose heart illness, and a modified synthetic ECG can be used to mimic some of these arrhythmias.

Emotions have significant importance in human life in learning, decision-making, daily interaction, and perception of the surrounding environment. Hence, it has become very essential to detect and recognize a person's emotional states and to build a connection between humans and computers. This process is called brain-computer interaction (BCI) and is a vast field of research in neuroscience. Hence, in the past few years, emotion recognition has gained adequate attention in the research community. In this thesis, an emotion recognition system is designed and analyzed using EEG signals. Several existing feature extraction techniques are studied, analyzed, and implemented to extract features from the EEG signals. An SVM classifier is used to classify the features into various emotional states. Four emotional states are detected, namely, happy, sad, anger, and relaxed state. The model is tested, and simulation results are presented with an interpretation. Furthermore, this study has mentioned and discussed the efficacy of the results achieved. The findings from this study could be beneficial in developing emotion-sensitive technologies, such as augmented modes of communication for severely disabled individuals who are unable to communicate their feelings directly.

The main objective of this thesis is to propose a localization method for a UAV using various computer vision and machine learning techniques. It plays a major role in planning the strategy for the flight, and acts as a navigational contingency method, in event of a GPS failure. The implementation of the algorithms employs high processing capabilities of the graphics processing unit, making it more efficient. The method involves the working of various neural networks, working in synergy to perform the localization. This thesis is a part of a collaborative project between The University of North Texas, Denton, USA, and the University of Windsor, Ontario, Canada. The localization has been divided into three phases namely object detection, recognition, and location estimation. Object detection and position estimation were discussed in this thesis while giving a brief understanding of the recognition. Further, future strategies to aid the UAV to complete the mission, in case of an eventuality, like the introduction of an EDGE server and wireless charging methods, was also given a brief introduction.

Electrocardiogram (ECG) is a biological signal that represents the heart's electrical activity. Interference from power lines introduces a frequency component of 50 to 60 Hz into the signal, which is the principal cause of ECG corruption. By using the Cadence Virtuoso Spectre circuit simulator and typical TSMC RF 180 nm CMOS technology, a notch filter was created to reduce powerline interference. The advantage of utilizing a notch filter for PLI is that noise at 60 Hz is completely eliminated without sacrificing any important information. Additionally, this study contains a MATLAB-based model for, which is used to compute the power spectral density for the obtained time-domain signal. By incorporating power spectral density into data gathering procedures, it is feasible to enhance data collection methodologies, construct models that appropriately account for observed power and aid in the removal of undesired components. NI LabVIEW is used to extract features. The advantage of ECG feature extraction is that it provides information that assists in the identification of cardiac rhythm issues, and gives information about the occurrence of heart attack. In this study, several patient data sets are utilized to extract characteristics and provide information regarding heart condition abnormalities.

It is known that quantum strategies change the range of possible payoffs for the players in the prisoner's dilemma. In this paper, we examine the effect of the degree of entanglement in determining the payoffs. When both players play quantum strategies, we show that the payoff for both players is unaffected by the entanglement value and it leads to a new Nash equilibrium.

This thesis consists of two projects in the field of machine learning. Previous research in the OSCAR UNT lab based on KMeans color quantization is further developed and applied to individual color channels and segmented input images to explore compression rates while still maintaining high output image quality. The second project implements a small-scale dual path network for image classifiaction utilizing the CIFAR-10 dataset containing 60,000 32x32 pixel images ranging across ten categories.

Wireless power transfer (WPT) is currently the most efficient way for transmission of power from one port to another, that is popularly used in various applications.This technique can change the previous energy utilization methods in various applications such as electronic devices, implanted medical devices, electrical vehicles and so forth.It mainly helps overcome the limitations of short battery life, limited storage, heavy weight, and high cost of batteries.This paper is based on the design of a transmitter and a receiver to achieve wireless power transfer for applications like optogenetic stimulation in rodents. With inductive coupling, a very high efficiency can be achieved between the transmitting and receiving coils of an antenna at small distances. When the transmitter and receiver are strongly coupled and are working at their resonant frequencies, the range of efficient WPT can be extended. In this work, the simulations are performed in HFSS at a resonating frequency of 13.56 MHz.A 4-port transmitter and a single-port planar receiver model are developed in HFSS, and the simulations are performed to graph the S parameters with a separation distance of 4cm. A Wilkinson power divider is designed using ADS to split the power from the four ports of the transmitter. The …

Currently, there is no equipment to measure the real-time fit of EHMR or N-95masks which are used in harsh environments. Improper fit of these EHMRs or N-95 masks exposes the personnel to hazardous environments. Surface acoustic wave (SAW) sensors have been around for few decades and are being used in various applications. In this work, real-time PM2.5 detection using passive wireless SAW sensors is presented. The design of meander antenna at 433MHz for wireless interrogation of SAW sensor using HFSS and ADS is also presented in this thesis. This works also includes the design of YZ-lithium niobate SAW sensor including COMSOL simulation.

Air corridors are an integral part of the advanced air mobility infrastructure. They are the virtual highways in the sky for transportation of people and cargo in the controlled airspace at an altitude of around 1000 ft. to 2000 ft. above the ground level. This paper presents fundamental insights into the design of air corridors with high operational efficiency as well as zero collisions. It begins with the definitions of air cube, skylane or track, intersection, vertiport, gate, and air corridor. Then, a multi-layered air corridor model is proposed. Traffic at intersections is analyzed in detail with examples of vehicles turning in different directions. The concept of capacity of an air corridor is introduced along with the nature of distribution of locations of vehicles in the air corridor and collision probability inside the corridor are discussed. Finally, the results of simulations of traffic flows are presented.

Low Earth orbit (LEO) satellite systems have been proposed as a resource for combating the challenges in 5G network coverage and expanding connectivity to a global realm. This research focuses on the current architecture of LEO satellite constellations, with an emphasis on satellite coverage, visibility patterns and coordination schemes. Key-elements of integrating LEO satellites into the eMBB component of 5G are presented and a breakdown of potential link channel characteristics and physical layer performance metrics are described. The produced information allows for a justified analysis on the conceptualized integration.

Broadly defined, group testing is the study of finding defective items in a large set. In the medical infection setting, that implies classifying each member of a population as infected or uninfected, while minimizing the total number of tests.

This thesis focuses on the recording challenges faced in biomedical systems. More specifically, the challenges in neural signal recording are explored. Instead of the typical synchronous ADC system, a level crossing ADC is detailed as it has gained recent interest for low-power biomedical systems. These systems take advantage of the time-sparse nature of the signals found in this application. A 10-bit design is presented to help capture the lower amplitude action potentials (APs) in neural signals. The design also achieves a full-scale bandwidth of 1.2 kHz, an ENOB of 9.81, a power consumption of 13.5 microwatts, operating at a supply voltage of 1.8 V. This design was simulated in Cadence using 180 nm CMOS technology.

Compressed sensing (CS) is an innovative approach of signal processing that facilitates sub-Nyquist processing of bio-signals, such as a neural signal, electrocardiogram (ECG), and electroencephalogram (EEG). This strategy can be used to lower the data rate to realize ultra-low-power performance, As the count of recording channels increases, data volume is increased resulting in impermissible transmitting power. This thesis work presents the implementation of a CMOS-based front-end design with the CS in the standard 180 nm CMOS process. A novel pseudo-random sequence generator is proposed, which consists of two different types of D flip-flops that are used for obtaining a completely random sequence. This thesis work also includes the (reverse electrowetting-on-dielectric) REWOD based energy harvesting model for self-powered bio-sensor which utilizes the electrical energy generated through the process of conversion of mechanical energy to electrical energy. This REWOD based energy harvesting model can be a good alternative to battery usage, particularly for the bio-wearable applications. The comparative analysis of the results generated for voltage, current and capacitance of the rough surface model is compared to that of results of planar surface REWOD.

The following thesis proposes solutions to problems in two main areas of focus, computer vision and machine learning. Chapter 2 utilizes traditional computer vision methods implemented in a novel manner to successfully identify overlays contained in broadcast footage. The remaining chapters explore machine learning algorithms and apply them in various manners to big data, multi-channel image data, and ECG data. L1 and L2 principal component analysis (PCA) algorithms are implemented and tested against each other in Python, providing a metric for future implementations. Selected algorithms from this set are then applied in conjunction with other methods to solve three distinct problems. The first problem is that of big data error detection, where PCA is effectively paired with statistical signal processing methods to create a weighted controlled algorithm. Problem 2 is an implementation of image fusion built to detect and remove noise from multispectral satellite imagery, that performs at a high level. The final problem examines ECG medical data classification. PCA is integrated into a neural network solution that achieves a small performance degradation while requiring less then 20% of the full data size.