Engineering Commons^™

Efficient Hardware Implementation Of Deep Learning Networks Based On The Convolutional Neural Network, Anaam Ansari

Engineering Ph.D. Theses

Image classification, speech processing, autonomous driving, and medical diagnosis have made the adoption of Deep Neural Networks (DNN) mainstream. Many deep networks such as AlexNet, GoogleNet, ResidualNet, MobileNet, YOLOv3 and Transformers have achieved immense success and popularity. However, implementing these deep and complex networks in hardware is a challenging feat. The growing demand of DNN applications in mobile devices and data centers have led the researchers to explore application specific hardware accelerators for DNNs. There have been numerous hardware and software based solutions to improve DNN throughput, latency, performance and accuracy. Any solution for hardware acceleration needs to optimize in …

Deep Learning-Based Low Complexity And High Efficiency Moving Object Detection Methods, Bingxin Hou

Engineering Ph.D. Theses

Moving object detection (MOD) is the process of extracting dynamic foreground content from the video frames, such as moving vehicles or pedestrians, while discarding the nonmoving background. It plays an essential role in computer vision field. The traditional methods meet difficulties when applied in complex scenarios, such as videos with illumination changes, shadows, night scenes,and dynamic backgrounds. Deep learning methods have been actively applied to moving object detection in recent years and demonstrated impressive results. However, many existing models render superior detection accuracy at the cost of high computational complexity and slow inference speed. This fact has hindered the development …

An Input Power-Aware Maximum Efficiency Tracking Technique For Energy Harvesting In Iot Applications, Sanad Fares Yousef Kawar

Engineering Ph.D. Theses

The Internet of Things (IoT) enables intelligent monitoring and management in many applications such as industrial and biomedical systems as well as environmental and infrastructure monitoring. As a result, IoT requires billions of wireless sensor network (WSN) nodes equipped with a microcontroller and transceiver. As many of these WSN nodes are off-grid and small-sized, their limited-capacity batteries need periodic replacement. To mitigate the high costs and challenges of these battery replacements, energy harvesting from ambient sources is vital to achieve energy-autonomous operation. Energy harvesting for WSNs is challenging because the available energy varies significantly with ambient conditions and in many …

A Drift-Resilient And Degeneracy-Aware Loop Closure Detection Method For Localization And Mapping In Perceptually-Degraded Environments, Kamak Ebadi

Engineering Ph.D. Theses

Enabling fully autonomous robots capable of navigating and exploring unknown and complex environments has been at the core of robotics research for several decades. Mobile robots rely on a model of the environment for functions like manipulation, collision avoidance and path planning. In GPS-denied and unknown environments where a prior map of the environment is not available, robots need to rely on the onboard sensing to obtain locally accurate maps to operate in their local environment. A global map of an unknown environment can be constructed from fusion of local maps of temporally or spatially distributed mobile robots in the …

Metamaterial Absorbers For Mitigating Unintentional Radiated Emissions, Ali Khoshniat

Engineering Ph.D. Theses

Unintentional radiated emission spikes are one of the causes of failure in electromagnetic compliance tests of high-speed systems. In this thesis, a new absorber solution for mitigating such emissions is proposed using the concept of metamaterial structures. The absorber is placed inside the high-speed system shield box to match the low (almost zero) impedance of the metal walls to the wave impedance of unwanted radiations. As a result, waves reflected from the shield box are attenuated which eventually reduces the emissions leaked outside of the box. The effectiveness of the proposed solution is demonstrated through simulations and experimental evaluations of …

A Flexible Bch Decoder For Flash Memory Systems Using Cascaded Bch Codes, Arul K. Subbiah

Engineering Ph.D. Theses

NAND ash memories are widely used in consumer electronics, such as tablets, personal computers, smartphones, and gaming systems. However, unlike other standard storage devices, these ash memories suffer from various random errors. In order to address these reliability issues, various error correction codes (ECC) are employed. Bose-Chaudhuri Hocquenghem (BCH) code is the most common ECC used to address the errors in modern ash memories. Because of the limitation of the realization of the BCH codes for more extensive error correction, the modern ash memory devices use Low-density parity-check (LDPC) codes for error correction scheme. The realization of the LDPC decoders …

Deep Learning For Recommender Systems, Travis Akira Ebesu

Engineering Ph.D. Theses

The widespread adoption of the Internet has led to an explosion in the number of choices available to consumers. Users begin to expect personalized content in modern E-commerce, entertainment and social media platforms. Recommender Systems (RS) provide a critical solution to this problem by maintaining user engagement and satisfaction with personalized content.

Traditional RS techniques are often linear limiting the expressivity required to model complex user-item interactions and require extensive handcrafted features from domain experts. Deep learning demonstrated significant breakthroughs in solving problems that have alluded the artificial intelligence community for many years advancing state-of-the-art results in domains such as …

Design And Measurement Of A Millimeter-Wave 2d Beam Switching Planar Antenna Array, Benjamin Horwath

Engineering Ph.D. Theses

A millimeter-wave 2-D beam switching microstrip patch antenna array excited by a 4x4 substrate integrated waveguide (SIW) Modified Butler Matrix is designed and experimentally evaluated in this thesis. A novel architecture is introduced for the Butler Matrix feed network to give designers a choice for phase shifter location to pursue a smaller circuit area. In addition, it enables the designer to control the BM phased outputs for achieving a set of desired 2-D beam directions, e.g., ϕ₀=45°, 135°, 225°, and 315° at θ₀=45°, with a passive beam switching network for a given array geometry. Full-wave simulation …

Quaternion Information Theoretic Learning Adaptive Algorithms For Nonlinear Adaptive, Carlo Safarian

Engineering Ph.D. Theses

Information Theoretic Learning (ITL) is gaining popularity for designing adaptive filters for a non-stationary or non-Gaussian environment [1] [2] . ITL cost functions such as the Minimum Error Entropy (MEE) have been applied to both linear and nonlinear adaptive filtering with better overall performance compared with the typical mean squared error (MSE) and least-squares type adaptive filtering, especially for nonlinear systems in higher-order statistic noise environments [3].

Quaternion valued data processing is beneficial in applications such as robotics and image processing, particularly for performing transformations in 3-dimensional space. Particularly the benefit for quaternion valued processing includes performing data transformations in …

Carbon Nanotube Ultracapacitor Characteristics And Cell Design, Antonis A. Orphanou

Engineering Ph.D. Theses

A model of carbon nanotube (CNT) ultracapacitor (CNU) as a high-performance energy storage device is developed based on simulations of electrolyte ion motions between cathode and anode. Using a molecular dynamics (MD) approach, the equilibrium positions of electrode charges interacting through Coulomb potential are determined, which in turn yield the equipotential surface and electric field associated with the capacitor. With an applied AC voltage, the current is computed from the nanotube and electrolyte particle distribution and interaction, resulting in a frequency-dependent CNU impedance. From the current and impedance profiles, the Nyquist and Cyclic Voltammetry plots are then extracted. Results of …

Carbon Nanotube Interconnects For End-Of-Roadmap Semiconductor Technology Nodes, Anshul A. Vyas

Engineering Ph.D. Theses

Advances in semiconductor technology due to aggressive downward scaling of on-chip feature sizes have led to rapid rises in resistivity and current density of interconnect conductors. As a result, current interconnect materials, Cu and W, are subject to performance and reliability constraints approaching or exceeding their physical limits. Therefore, alternative materials such as nanocarbons, metal silicides, and Ag nanowires are actively considered as potential replacements to meet such constraints. Among nanocarbons, carbon nanotube (CNT) is among the leading replacement candidate for on-chip interconnect vias due to its high aspect-ratio nanostructure and superior currentcarrying capacity to those of Cu, W, and …

Scalable Speech Coding For Ip Networks, Koji Seto

Engineering Ph.D. Theses

The emergence of Voice over Internet Protocol (VoIP) has posed new challenges to the development of speech codecs. The key issue of transporting real-time voice packet over IP networks is the lack of guarantee for reasonable speech quality due to packet delay or loss.

Most of the widely used narrowband codecs depend on the Code Excited Linear Prediction (CELP) coding technique. The CELP technique utilizes the long-term prediction across the frame boundaries and therefore causes error propagation in the case of packet loss and need to transmit redundant information in order to mitigate the problem. The internet Low Bit-rate Codec …

Power Reductions With Energy Recovery Using Resonant Topologies, Ignatius S.A. Bezzam

Engineering Ph.D. Theses

The problem of power densities in system-on-chips (SoCs) and processors has become more exacerbated recently, resulting in high cooling costs and reliability issues. One of the largest components of power consumption is the low skew clock distribution network (CDN), driving large load capacitance. This can consume as much as 70% of the total dynamic power that is lost as heat, needing elaborate sensing and cooling mechanisms. To mitigate this, resonant clocking has been utilized in several applications over the past decade. An improved energy recovering reconfigurable generalized series resonance (GSR) solution with all the critical support circuitry is developed in …