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.

A modified, single error-correcting, and double error detecting Hamming code, hereafter referred to as modified SEC-DED Hamming code, is proposed in this research. The code requires fewer logic gates to implement than the SEC-DED Hamming code. Also, unlike the popular Hsiao's code, the proposed code can determine the error in the received word from its syndrome location in the parity check matrix. A detailed analysis of the area and power utilization by the encoder and decoder circuits of the modified SEC-DED Hamming code is also discussed. Results demonstrate that this code is an excellent alternative to Hsiao's code as the area and power values are very similar. In addition, the ability to locate the error in the received word from its syndrome is also of particular interest. Primitive polynomials play a crucial role in the hardware realizations for error-correcting codes. This research describes an implementation of a scalable primitive polynomial circuit with coefficients in GF(2). The standard cell area and power values for various degrees of the circuit are analyzed. The physical design of a degree 6 primitive polynomial computation circuit is also provided. In addition to the codes, a background of the already existing SPX GCD computation algorithm is …

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.

Secure multiparty computation refers to the problem where a number of users wish to securely compute a function on their inputs without revealing any unnecessary information. This dissertation focuses on the fundamental limits of secure summation under different constraints. We first focus on the minimal model of secure computation, in which two users each hold an input and wish to securely compute a function of their inputs at the server. We propose a novel scheme base on the algebraic structure of finite field and modulo ring of integers. Then we extend the minimal model of secure computation, in which K users wish to securely compute the sum of their inputs at the server. We prove a folklore result on the limits of communication cost and randomness cost. Then we characterized the optimal communication cost with user dropouts constraint, when some users may lose connection to the server and the server wishes to compute the sum of remaining inputs. Next, we characterize the optimal communication and randomness cost for symmetric groupwise keys and find the feasibility condition for arbitrary groupwise keys. Last, we study the secure summation with user selection, such that the server may select any subset of users to …

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.

Among various energy harvesting technologies reverse electrowetting-on-dielectric energy harvesting (REWOD) has been proved to harvest energy from low frequency motion such as many human motion activities (e.g. walking, running, jogging etc.). Voltage rectification and DC-DC boosting of low magnitude AC voltage from REWOD can be used to reliably self-power the wearable sensors. In this work, a commercial component-based rectifier and DC-DC converter is designed and experimentally verified, for further miniaturization standard 180 nm CMOS process is used to design the rectifier and the DC-DC boost converter.This work also includes the MATLAB based model for REWOD energy harvester for various REWOD models. In REWOD energy harvesting, a mechanical input during the motion causes the electrolyte placed in between two dissimilar electrodes to squeeze back and forth thereby periodically changing the effective interfacial area, hence generating alternating current. The alternating current is given to the rectifier design. There is no realistic model that has been developed yet for this technique. Thereby, a MATLAB based REWOD model is developed for the realistic simulation of the REWOD phenomenon. In the work, a comparison of different REWOD models such as planar surface, rough surface and porous models are performed demonstrating the variations in capacitance, current …

This dissertation focuses on the path planning of unmanned aerial vehicle (UAV) swarms under distributed and hybrid control scenarios. It presents two such models and analyzes them both from theory and practice. In the first method, a distributed formation control strategy for UAV swarm based on consensus law is presented. This model makes use of the fundamental concepts of leader-follower structure, social potential functions, and algebraic graph theory to jointly address flocking and de-confliction in the formation control problem. The impact of network topology on formation control is analyzed. It is shown that the degree distribution of the network representing the multi-agent system defines the rate at which formation is attained. Conditions for convergence and stability are derived. In the second method, a hybrid framework for path planning and coverage area by UAV swarms is presented. This strategy significantly improves the current labor-intensive and resource-constraint operations in aquaculture farms. To monitor the farms periodically, an optimized back-and-forth flight path based on the Shamos algorithm is utilized. A trajectory tracking strategy for UAV swarms under uncertain wind conditions is presented.

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.

This dissertation is devoted to the development of novel devices for optoelectronic and photovoltaic applications using the promise of inkjet printing with two-dimensional (2D) materials. A systematic approach toward the characterization of the liquid exfoliated 2D inks comprising of graphene, molybdenum disulfide (MoS2), tungsten diselenide (WSe2), and 2D perovskites is discussed at depth. In the first study, the biocompatibility of 2D materials -- graphene and MoS2 -- that were drop cast onto flexible PET and polyimide substrates using mouse embryonic fibroblast (STO) and human esophageal fibroblast (HEF) cell lines, was explored. The polyimide samples for both STO and HEF showed high biocompatibility with a cell survival rate of up to ~ 98% and a confluence rate of 70-98%. An inkjet printed, biocompatible, heterostructure photodetector was constructed using inks of photo-active MoS2 and electrically conducting graphene, which facilitated charge collection of the photocarriers. The importance of such devices stems from their potential utility in age-related-macular degeneration (AMD), which is a condition where the photosensitive retinal tissue degrades with aging, eventually compromising vision. The biocompatible inkjet printed 2D heterojunction devices were photoresponsive to broadband incoming radiation in the visible regime, and the photocurrent scaled proportionally with the incident light intensity, exhibiting a …

With the rapid development of wireless communication technology, the demands for higher communication rates are increasing. Higher communication rate corresponds to higher DoF. Interference alignment, which is an emerging interference management technique, is able to substantially increase the DoF of wireless communication systems. This thesis mainly studies the delay-based interference alignment technique. The key problem lies in the design of the transmission scheme and the appropriate allocation of the propagation delay, so as to achieve the desired DoF of different wireless networks. In addition, through delay-based interference alignment, the achievability of extreme points of the DoF region of different wireless networks can be proved.

The new developments in coding theory research have revolutionized the application of coding to practical systems. Low-Density Parity Check (LDPC) codes form a class of Shannon limit approaching codes opted for digital communication systems that require high reliability. This thesis investigates the underlying relationship between the spectral properties of the parity check matrix and LDPC decoding convergence. The bit error rate of an LDPC code is plotted for the parity check matrix that has different Second Smallest Eigenvalue Modulus (SSEM) of its corresponding Laplacian matrix. It is found that for a given (n,k) LDPC code, large SSEM has better error floor performance than low SSEM. The value of SSEM decreases as the sparseness in a parity-check matrix is increased. It was also found from the simulation that long LDPC codes have better error floor performance than short codes. This thesis outlines an approach to analyze LDPC decoding based on the eigenvalue analysis of the corresponding parity check matrix.

In light of incidents and concerns regarding the vulnerability of the global positioning system (GPS), the main purpose of the thesis is to look at alternative systems for radio guidance and to put up a serious study on such alternatives with receive and transmit antenna. There is also the need to design such antennas with multiple frequencies to offer robustness in the unlikely event that such adversarial attacks on the GPS happen. The basis on which such alternative antennas are designed is a slotted microstrip. The characteristics of the slot or slots on the microstrip are analyzed by mapping their exact locations on the patch and then noting the resultant center frequencies, the return losses, and the bandwidth. The activities associated with this also focus on the design, fabrication, validation, and characterization of one or more slotted antennas prototypes. The measurement of the antenna prototypes does confirm several frequencies that coexist to see applications, in aeronautical radionavigation, fixed-mobile radionavigation, and radiolocation. The antennas could also feature in a wide-area augmentation system (WAAS), satellite ground link system (SGLS) as well as in surveillance and precision approach radars. Some variations of the antenna are deployed in the areas of law enforcement, surveillance, …

Security and privacy are of paramount importance in the modern information age. Secure multi-party computation and private information retrieval are canonical and representative problems in cryptography that capture the key challenges in understanding the fundamentals of security and privacy. In this dissertation, we use information theoretic tools to tackle these two classical cryptographic primitives. In the first part, we consider the secure multi-party computation problem, where multiple users, each holding an independent message, wish to compute a function on the messages without revealing any additional information. We present an efficient protocol in terms of randomness cost to securely compute a vector linear function. In the second part, we discuss the symmetric private information retrieval problem, where a user wishes to retrieve one message from a number of replicated databases while keeping the desired message index a secret from each individual database. Further, the user learns nothing about the other messages. We present an optimal protocol that achieves the minimum upload cost for symmetric private information retrieval, i.e., the queries sent from the user to the databases have the minimum number of bits.

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.

By organically fusing the methods of convolutional neural network (CNN) and recurrent neural network (RNN), this dissertation focuses on the application of optical character recognition and image classification processing. The first part of this dissertation presents an end-to-end novel receipt recognition system for capturing effective information from receipts (CEIR). The main contributions of this research part are divided into three parts. First, this research develops a preprocessing method for receipt images. Second, the modified connectionist text proposal network is introduced to execute text detection. Third, the CEIR combines the convolutional recurrent neural network with the connectionist temporal classification with maximum entropy regularization as a loss function to update the weights in networks and extract the characters from receipt. The CEIR system is validated with the scanned receipts optical character recognition and information extraction (SROIE) database. Furthermore, the CEIR system has strong robustness and can be extended to a variety of different scenarios beyond receipts. For the convolutional recurrent structure application of land use image classification, this dissertation comes up with a novel deep learning model for land use classification, the convolutional recurrent land use classifier (CRLUC), which further improves the accuracy in classifying remote sensing land use images. Besides, the …

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.

Sight is essential in our daily tasks. Compensatory senses have been used for centuries by visually impaired individuals to navigate independently. The help of technology can minimize some challenges for visually impaired individuals. Assistive navigation technologies facilitate the pathfinding and tracing in indoor scenarios. Different modules are added to assistive navigation technologies to warn about the obstacles not only on the ground but about hanging objects. In this work, we attempt to explore new methods to assist visually impaired individuals in navigating independently in an indoor scenario. We employed a location estimation algorithm based on the fingerprinting method to estimate the initial location of the user. We mitigate the error of estimation with particle filter. The shortest path has been calculated with an A* algorithm. To provide the user with an accident-free experiment, we employed an obstacle avoidance algorithm capable of warning the users about the potential hazards. Finally, to provide an effective means of communication with the user, we employed text-to-speech and speech recognition algorithms. The main contribution of this work is to glue these modules together efficiently and affordably.

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.

In this work, we explored one material from the broad family of 2D semiconductors, namely WSe2 to serve as an enabler for advanced, low-power, high-performance nanoelectronics and optoelectronic devices. A 2D WSe2 based field-effect-transistor (FET) was designed and fabricated using electron-beam lithography, that revealed an ultra-high mobility of ~ 625 cm2/V-s, with tunable charge transport behavior in the WSe2 channel, making it a promising candidate for high speed Si-based complimentary-metal-oxide-semiconductor (CMOS) technology. Furthermore, optoelectronic properties in 2D WSe2 based photodetectors and 2D WSe2/2D MoS2 based p-n junction diodes were also analyzed, where the photoresponsivity R and external quantum efficiency were exceptional. The monolayer WSe2 based photodetector, fabricated with Al metal contacts, showed a high R ~502 AW-1 under white light illumination. The EQE was also found to vary from 2.74×101 % - 4.02×103 % within the 400 nm -1100 nm spectral range of the tunable laser source. The interfacial metal-2D WSe2 junction characteristics, which promotes the use of such devices for end-use optoelectronics and quantum scale systems, were also studied and the interfacial stated density Dit in Al/2D WSe2 junction was computed to be the lowest reported to date ~ 3.45×1012 cm-2 eV-1. We also examined the large exciton binding …

A wearable body temperature sensor would allow for early detection of fever or infection, as well as frequent and accurate hassle-free recording. This thesis explores the design of a body-temperature-sensing device inkjet-printed on a flexible substrate. All structures were first modeled by first-principles, theoretical calculations, and then simulated in HFSS. A variety of planar square inductor geometries were studied before selecting an optimal design. The designs were fabricated using multiple techniques and compared to the simulation results. It was determined that inductance must be carefully measured and documented to ensure good functionality. The same is true for parallel-plate and interdigitated capacitors. While inductance remains relatively constant with temperature, the capacitance of the device with a temperature-sensitive dielectric layer will result in a shift in the resonant frequency as environmental or ambient temperature changes. This resonant frequency can be wirelessly detected, with no battery required for the sensing device, from which the temperature can be deduced. From this work, the optimized version of the design comprises of conductive silver in with a temperature-sensitive graphene oxide layer, intended for inkjet-printing on flexible polyimide substrates. Graphene oxide demonstrates a high dielectric permittivity with good sensing capabilities and high accuracy. This work pushes the …

In this work, we study the properties and device applications of MoS2, black phosphorus, MoOx, and NbSe2. We first start with the design, fabrication, and characterization of ultra-high responsivity photodetectors based on mesoscopic multilayer MoS2. The device architecture is comprised of a metal-semiconductor-metal (MSM) photodetector, where Mo was used as the contact metal to suspended MoS2 membranes. The dominant photocurrent mechanism was determined to be the photoconductive effect, while a contribution from the photogating effect was also noted from trap-states that yielded a wide spectral photoresponse from UV-to-IR with an external quantum efficiency (EQE) ~ 104. From time-resolved photocurrent measurements, a fast decay time and response time were obtained with a stream of incoming ON/OFF white light pulses. Another interesting semiconductor 2D material that has attracted special attention due to its small bandgap and ultra-high hole mobility is the black phosphorus. An analysis of the optoelectronic properties and photocurrent generation mechanisms in two-dimensional (2D) multilayer crystallites of black phosphorus (BP) was conducted from 350 K down to cryogenic temperatures using a broad-band white light source. The Mo-BP interface yielded a low Schottky barrier "φ" _"SB" ~ -28.3 meV and a high photoresponsivity R of ~ 2.43 x 105 A/W at …

Although wireless sensor networks have been successfully used for environmental monitoring, one of the major challenges that this technology has been facing is supplying continuous and reliable electrical power during long-term field deployment. Batteries require repetitive visits to the deployment site to replace them once discharged; admittedly, they can be recharged from solar panels, but this only works in open areas where solar radiation is unrestricted. This dissertation introduces a novel approach to design and implement a reliable efficient solar energy harvester to continuously, and autonomously, provide power to wireless sensor nodes for long-term applications. The system uses supercapacitors charged by a solar panel and is designed to reduce power consumption to very low levels. Field tests were conducted for more than a year of continuous operation and under a variety of conditions, including areas under dense foliage. The resulting long-term field data demonstrates the feasibility and sustainability of the harvester system for challenging applications. In addition, we analyzed solar radiation data and supercapacitor charging behavior and showed that the harvester system can operate battery free, running on the power provided by supercapacitors. A battery is included only for backup in case the supercapacitor storage fails. The proposed approach provides …

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.

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.