Browsing by Author "Gan, Yu"
Now showing 1 - 18 of 18
Results Per Page
Sort Options
Item A CNN-Based Microwave Imaging System for Detecting Watermelon Ripeness(IEEE, 2025-01-28) Choffin, Zachary M.; Kong, Lingyang; Gan, Yu; Jeong, NathanThe integration of a non-invasive microwave imaging system with a machine learning algorithm could improve food quality and food safety. In this paper, a S- and C-band microwave imaging system that utilizes DAS (Delay and Sum) beamforming with an automated high-frequency switching network is built to scan watermelons and determine their ripeness. A total of 288 images were collected from eight different watermelons varying the height and angle of capture. A convolutional neural network (CNN) was employed to assess the ripeness level, which was determined by analyzing the Brix sugar content. The results show 86% accuracy for ripeness classification in three fold cross validation. This novel approach demonstrates the potential of combining microwave imaging with machine learning for non-destructive food quality assessment, offering a scalable and reliable tool for real-time evaluation of fruit ripeness and quality.Item American Sign Language Recognition Using Adversarial Learning in a Multi-Frequency RF Sensor Network(University of Alabama Libraries, 2020) Macks, Trevor F; Gurbuz, Sevgi Z; University of Alabama TuscaloosaHuman computer interaction (HCI) based technologies, such as Alexa or Siri, have become increasingly prevalent in the daily lives of American citizens. However, access to these technologies is gated by the need to communicate with spoken commands, which subsequently precludes the Deaf community from benefitting from the quality of life improvements they provide. Current approaches to providing HCI technologies or ASL conversion largely rely on video and image processing techniques, haptic gloves, and wifi based systems. However, wearables restrict users from engaging in their normal daily activities, while video raises privacy concerns. To help ASL compatible HCI technology, we propose a multi-frequency RF sensing network for the recognition of a basic lexicon of signs. When validating the network on a daily activities dataset, we had good performance, with classification accuracy's around 90% or higher. For ASL data, we have 2 datasets: native and imitation. The native dataset is small, and was collected from Deaf individuals. Our imitation dataset is larger, and was collected from hearing individuals prompted by copysigning videos. We show that imitation data cannot be used in lieu of native ASL data for training and benchmarking classifiers because the two datasets possess disparate distributions in feature space. Alternatively, we investigate adversarial learning as a means for mitigating the challenge of insufficient training data. Cross frequency training is one option for augmenting the training dataset which suffers from severe performance degradation when data from one frequency is used to pre-train a network for classification of data at another frequency. We show that data synthesized using Generative Adversarial Networks (GANs) can be used to reduce but not completely eliminate cross-frequency training degradation. An auxiliary conditional generative adversarial network (ACGAN) with kinematic sifting is used to augment and classify human activity data and recognize ASL signs. While the proposed network performed well with daily activities, its performance could not be adequately validated on ASL data due to sparsity of native ASL data and statistical inconsistencies of imitation signing data. Future directions for overcoming these challenges and extending the proposed techniques to ASL recognition are discussed.Item Artificial Intelligence and Machine Learning for Control and Operation of Electric Vehicles and Machine Drives(University of Alabama Libraries, 2021) Dong, Weizhen; Li, Shuhui; University of Alabama TuscaloosaMotor drive and charging system with batteries are two major parts in an electric vehicle (EV) powertrain system. This dissertation investigates the artificial intelligence-based control and operation of EV machine drives and the charging systems.There are several major challenges related the EV motor drive control such as machine parameter variations, magnetic saturations, accurate torque control, and optimal and efficient operation considering copper loss and iron loss. Regarding the charging and discharging control with DC/DC converters, the stable and robust voltage regulation under disturbances is required. The issues of how to smoothly handle the current/voltage constraints and the power limit still remain. This dissertation presents a novel machine learning strategy based on a neural network (NN) to achieve MTPA, flux-weakening, and MTPV for the most efficient IPM torque control over its full speed operating range. The NN is trained offline by using the LMBP (Levenberg-Marquardt backpropagation) algorithm, which avoids the disadvantages associated with the online NN training. A special technique is developed to generate NN training data, that is particularly suitable and favorable, to develop a high-performance NN-based IPM torque control system, and the impact of variable motor parameters is embedded into the NN system development and training. IPM machine modeling and parameter estimation are important for the controller design of high-efficient and high-performance motor drives. The accuracy of the magnetic modeling is a challenge dur to the magnetic saturation, cross saturation, iron loss, and temperature variations. The proposed ANN-based modeling method can capture the nonlinear areas of the model and generate accurate dq-axis flux linkages with saturation and iron loss considered. For the vehicle to grid (V2G) and vehicle to home (V2H) applications, the battery not only can be charged but also can provide power back to the load and systems through DC/DC converters. The ANN controller presented in this dissertation has a strong ability to track rapidly changing reference commands, maintain stable output voltage for a variable load, and manage maximum duty-ratio and current constraints properly. The presented control algorithm also has the ability of power sharing based on DG capabilities for DC microgrid applications.Item Artificial Intelligence for Building Energy Management in the Electricity Market and Transmission Power Flow Planning(University of Alabama Libraries, 2021) Gao, Yixiang; Li, Shuhui; University of Alabama TuscaloosaDue to the high uncertainty of building loads and customer comfort demands and extremely nonlinear building thermal characteristics, developing an effective building energy management (BEM) technology is facing great challenges. This dissertation focuses on building energy management from the day-ahead and real-time planning perspectives in the electricity market. In the day-ahead planning, this dissertation presents price-sensitive demand response strategies for smart buildings by regulating their controllable loads to minimize building electricity costs and flatten the net buildings' loads. The learning-based HVAC model and the detailed physics-based non-HVAC model are then applied to an optimization problem to determine the optimal management scheduling of building loads based on day-ahead electricity price. In addition, this dissertation proposes an hourly decoupled AC/DC power flow approach for the day-ahead planning of multi-terminal HVDC systems. The proposed method simplifies the power flow computation of multi-terminal HVDC systems while accurately reflecting the operation and control characteristics of VSC (voltage source converter) stations in an HVDC network. In real-time planning, an optimization problem in a 5-minute time frame is proposed to manage real-time building energy consumption uncertainties based on the real-time clearing price and balance the real-time deviations from the building energy consumption negotiated in the day-ahead market. To help regulate power system real-time frequency fluctuation, an economic and hierarchical control approach for multi-thermal-zone buildings is developed to participate in the ancillary service market of the electric power systems. The energy consumption of variable speed drive (VSD) fans in the HVAC system is controlled up and down to follow dynamic auto-generation control (AGC) signals from the power system operators or control centers. A decoupling method is developed to balance the power system frequency regulation requirement in 5 seconds and real-time building energy planning developed above in 5 minutes. In this dissertation, computer simulation models for building and grid integration from different power systems planning perspectives are developed. All the proposed methods are studied theoretically and evaluated via computer simulations, based on which results and conclusions for each is obtained and reported in the dissertation.Item Conjugate operator on variable harmonic Bergman space(University of Alabama Libraries, 2020) Wang, Xuan; Ferguson, Timothy; University of Alabama TuscaloosaComplex analytic functions have astonishing and amazing properties. Their real parts and imaginary parts are deeply connected by the Cauchy-Riemann equations. It is natural to ask if we obtain some information about the real part, what can we conclude about the imaginary part, which is called the harmonic conjugate of the real part? Treating the relationship as an operation, the question becomes how well behaved is the harmonic conjugate operator? In this paper, by modifying some classical methods in constant exponent Hardy and Bergman spaces and developing new ways for the modern variable exponent spaces, we will study the harmonic conjugate operator on variable exponent Bergman spaces and prove that the operator is bounded when the exponent has positive minimum and finite maximum and satisfies the log-Holder condition.Item High Probability of Hiding (HPH) and Quality of Autonomy (QoA) Orientated 3D Transformative Intelligent Routing for Airborne Networks(University of Alabama Libraries, 2020) Zhang, Lin; Hu, Fei; Kumar, Sunil; University of Alabama TuscaloosaAirborne networks (ANs), such as networks of aircraft or flying drones, are finding use in various applications. Research on ANs is performed in this article from a communication protocol design perspective. Multiple flying objects with same or similar tasks generally form a flocking network where each object is considered as an airborne node. High probability of hiding (HPH)-oriented communication, which aims to avoid easy signal detection by adversaries, is thus very important for a robust AN. To protect the data of wireless communications, HPH requires to hide the very existence of radio signals [7]. Chapter 2 focuses on the robust routing protocols in ANs. Particularly, an AN equipped with the latest antenna technology, called multi-beam directional antenna (MBDA), is mainly considered. MBDA allows the simultaneous packet delivery in multiple directions without RF interference among the antenna beams. We have found that MBDAs can actually bring new opportunities to enhance the popularly used AN routing scheme, i.e., optimized link state routing protocol (OLSR). In particular, MBDAs enable OLSR to better achieve HPH, a critical requirement in many applications with adversary nodes nearby which try to eavesdrop the signals. Chapter 3 aims to solve a challenging issue in the ANs: in a three-dimensional (3D) airborne network which performs flocking operations (i.e., changing formations from time to time), how do we efficiently establish/maintain one or multiple paths between region regions (i.e., groups of airborne nodes with location proximity and task similarity), during the dynamic airborne nodes flocking process? A distributed routing scheme is desired that only uses localized message exchange (among 1-hop neighbors) without GPS position information. In Chapter 4, a task-adaptive, quality of autonomy (QoA)-based band routing scheme is analyzed. Adaptive Batch Coding (ABC)-based transport and routing layer co-design realizes smart inter-region routing for high priority task-command traffic. Scored time-delay embedding (STDE) technique provides a way for finding out and representing the time series’ periodicity quantitatively, which is suitable for airborne node stability estimation. Multi-time-granularity prediction (MTGP)-based band routing scheme is utilized to make the airborne communication support QoA metrics. Early backup (detour) node/path setup is achieved by the prediction results.Item Inpainting for Saturation Artifacts in Optical Coherence Tomography Using Dictionary-Based Sparse Representation(IEEE, 2021) Liu, Hongshan; Cao, Shengting; Ling, Yuye; Gan, Yu; University of Alabama Tuscaloosa; Shanghai Jiao Tong UniversitySaturation artifacts in optical coherence tomography (OCT) occur when received signal exceeds the dynamic range of spectrometer. Saturation artifact shows a streaking pattern and could impact the quality of OCT images, leading to inaccurate medical diagnosis. In this paper, we automatically localize saturation artifacts and propose an artifact correction method via inpainting. We adopt a dictionary-based sparse representation scheme for inpainting. Experimental results demonstrate that, in both case of synthetic artifacts and real artifacts, our method outperforms interpolation method and Euler's elastica method in both qualitative and quantitative results. The generic dictionary offers similar image quality when applied to tissue samples which are excluded from dictionary training. This method may have the potential to be widely used in a variety of OCT images for the localization and inpainting of the saturation artifacts.Item Intelligent Treadmill Control and Holographic Rendering for Accessible Rehabilitation(University of Alabama Libraries, 2025) Cao, Shengting; Hu, FeiThere is a growing need in the medical rehabilitation market due to the increasing elderly population. More than half of the patients are outpatients who require commuting from home to nursing facilities. However, the geographic distribution of these facilities is highly imbalanced, with states like Texas and California housing the largest number, while many rural and remote areas face a shortage. This disparity creates a significant burden for patients who require continuous rehabilitation but struggle with long commutes. Addressing this gap in rehabilitation accessibility is the central focus of this dissertation. This work approaches the problem from two complementary directions: (1) reducing the cost of home-based rehabilitation equipment, and (2) advancing telerehabilitation technologies. The first part of the dissertation introduces an intelligent treadmill control system that enables a single-belt treadmill to function like a split-belt treadmill, thereby providing a cost-effective solution for post-stroke gait rehabilitation at home. This system integrates real-time gait classification models and adaptive speed control to simulate split-belt dynamics without requiring specialized hardware. The second part explores novel telerehabilitation solutions. We adopted advanced neural rendering techniques to build a low-cost 3D patient reconstruction pipeline to enhance remote patient monitoring and engagement. Through these innovations, this dissertation contributes to making rehabilitation more accessible, affordable, and effective, particularly for patients in underserved areas. The proposed solutions aim to bridge the gap between clinical rehabilitation and home-based care, ultimately improving patient outcomes and reducing healthcare disparities.Item Leveraging Temporal Information for Fast Object Detection in High-Resolution Videos(University of Alabama Libraries, 2021) Adreon, Colin; Gan, Yu; University of Alabama TuscaloosaDetecting objects in high-resolution videos in real-time has proven extremelydifficult. The large size of high-resolution images makes traditional object- detection methods impractical within the short time period between video frames. Previous approaches to this problem have relied on techniques which select re- gions to analyze within a frame through pyramid pooling and attention pipelin- ing. We propose a novel approach which uses historical location information from earlier frames to inform decisions relating to specific regions in later frames. When run on a dataset of 4k videos, this approach has shown significant improve- ments in temporal efficiency without reducing accuracy over both attention- based methods and more naı̈ve approaches. At lower frame rates, this algorithm is able to process high-resolution video data in real time and can be used to monitor video camera footage without human intervention.Item Machine learning enhanced 5G vehicle-to-everything (V2X) communication networks with millimeter-waves and terahertz links(University of Alabama Libraries, 2020) Rasheed, Iftikhar; Hu, Fei; University of Alabama TuscaloosaWith the incoming of 5G communications, Vehicular Networks have the hope to achieve ultra-high data transmission rate with extremely low end-to-end delay. However, the dynamic nature of transportation traffic and increased data bandwidth demands are the major obstacles to achieve high transmission rate in Vehicular-to-Anything (V2X) Networks. To overcome these obstacles, this work presents a novel Software Defined Networking(SDN)-controlled and Cognitive Radio (CR)-enabled V2X routing approach to achieve ultra-high data rate, by using predictive V2X routing that supports the intelligent switching between two 5G technologies: millimeter-wave (mmWave) and terahertz (THz). To improve the network management, Road Side units (RSUs) are used to segregate the V2X network into different clusters. Stability-aware clustering (SAC) scheme is also used for cluster formations. The proposed intelligent V2X network is based on three features enabled machine learning approach: (1) To predict future 3D positions of the vehicles in the Cluster Heads (CHs) using Deep Neural Network with Extended Kalman Filter (DNN-EKF) algorithm for real-time, high-resolution prediction. (2) For THz communications, 0.3 THz to 3 THz band is selected for short-distance super-fast data transmissions. The THz band detection is performed by the CR-enabled Road Side Units (cRSUs). A Genetic Algorithm (GA)-based Improved Fruit Fly (GA-IFF) scheme is proposed to achieve an optimal route selection in THz communications. (3) In mmWave based V2X communications, optimal beam selection is performed by the multi-type2 fuzzy inference system (M-T2FIS). By using these three intelligent designs approaches, we are able to achieve ultrahigh data rate and minimized transmission delay for short-range (in THz bands) and middle-range (in mmWave) communications. Finally, the proposed SDN-controlled, CR-enabled V2X Network is modeled and tested for performance evaluations with the metrics of delivery ratio, routing delay, protocol overhead, and data rate. This work consists of effective cluster formation, intelligent switching, optimal path selection, and optimal beam selection. And it provides high data rate with lower latency and better reliability which is very much necessary for V2X communications.Item Machine Learning-Driven Intelligent Shoe-Based Wearable System for Human Health Enhancement(University of Alabama Libraries, 2024) Choffin, Zachary Michael; Jeong, NathanIn the last few decades, biomechanical research has made significant advancements in understanding human movement and developing techniques to improve individuals' everyday life. However, risks such as falls, musculoskeletal injuries, and improper biomechanics are still prevalent across all age groups and populations. While research has expanded the understanding of these risks, there is an ever-growing need for prevention strategies that can be seamlessly integrated into daily life. The development of intelligent wearable shoe sensor systems can open the door to transform the way we approach biomedical health and movement analysis. These discreet, unobtrusive devices can continuously monitor biomechanical data like gait patterns, joint angles, center of pressure, and balance, providing a way for individuals to mitigate risks and for doctors to have continuous data on how patients progress with treatment. This dissertation examines the development and use cases of the intelligent shoe-based wearable in biomechanical applications. It proposes an all-in-one insole capable of being discreet and low-cost to measure pressure across a user's foot and digitize this data for use in machine learning applications. A lower body joint angle detection model utilizing inverse dynamics is proposed. A method of classifying human motion and uniquely identifying individuals utilizing intelligent shoe-based wearable data is also proposed. Finally, a framework for utilizing center of pressure across the foot to correlate to a 2-dimensional center of balance is presented as an alternative to traditional IMU-based systems.Item Measurement and Analysis of Cigarette Smoke Exposure and Smoking Behavior Using Wearable Sensors(University of Alabama Libraries, 2021) Belsare, Prajakta; Sazonov, Edward; University of Alabama TuscaloosaCigarette smoking is the most prevalent cause of preventable deaths in the whole world. There are hundreds of toxins in a single cigarette that can have harmful effects on both active and passive smokers. Researchers think it is important to understand and provide accurate information regarding daily cigarette smoking and smoke exposure to understand the health impacts of cigarette consumption. There are many tools available for estimation of daily cigarette consumption such as self-report, biomarkers of cigarette smoke exposure, puff topography devices, and, recently, wearable sensors. However, these methods have few limitations, such as recall biases or digit preference in self-reporting. Biomarkers are objective and accurate, but they are expensive and not feasible for monitoring daily consumption. Puff-topography devices can provide puffing and cigarette consumption information but fail to report the post-puff information, such as duration of cigarette smoke holding in the lungs. Research shows wearable sensors can objectively and automatically detect cigarette smoking in the free-living environment. However, they are limited to detecting the number of cigarettes, the number of puffs, duration of puff, or duration of cigarette smoking. None of the methods available to date can identify the post-puff information such as depth of inhalation, smoke holding duration, etc. This information is vital in understanding the detailed smoking behavior of an individual smoker. Thus, there was a need for the development of a reliable method for extracting the puffing and post-puffing information of daily cigarette consumption of individual smoker. This dissertation proposes the use of breathing signal for extracting smoke exposure metrics. This dissertation also proposes the development of deep learning architecture for monitoring cigarette smoking in free-living; and signal processing/pattern recognition methods for extracting post-puff information. The main accomplishments of this dissertation are (a) review of existing methods for monitoring cigarette smoking and measurement of cigarette smoke exposure; (b) development of a novel algorithm named RSEM (Respiratory Smoke Exposure Metrics) for extracting puffing and post-puffing information ( such as puff duration, inhale-exhale duration and volume, volume over time, smoke hold duration, inter-puff interval) from breathing signal. The proposed algorithm provided measures previously unavailable in research; (c) establishing a relationship between smoke exposure metrics computed using RSEM algorithm and the biomarkers of smoke exposure expire CO and Cotinine level; (d) development of DeepPuff algorithm for automatic identification of smoking inhalation in the free-living environment; (e) analysis of cigarette smoking in the free-living environment and the effects of using puff topography devices on the number of puffs, smoking duration, puff duration, inhale-exhale duration, inhale-exhale volume, smoke hold duration, and inter-puff interval.Item Microwave imaging for watermelon maturity determination(Elsevier, 2023) Garvin, Joe; Abushakra, Feras; Choffin, Zachary; Shiver, Bayley; Gan, Yu; Kong, Lingyan; Jeong, Nathan; University of Alabama Tuscaloosa; Stevens Institute of TechnologyMicrowave imaging technology is a useful method often applied in medical diagnosis and can be used by the food industry to ensure food safety and quality. For fruit, ripeness is the primary characteristic which determines quality for the consumer. This paper proposes a novel microwave imaging system to determine the ripeness of watermelon as a proof of concept. The design employs a circular array with 10 Coplanar Vivaldi antennas offering wide bandwidth, high gain, and high efficiency. S-parameters between antennas are collected quickly via automated channel switching for fast image generation. Eight different watermelon samples of varying ripeness, type, dimensions, and origin are scanned and imaged. Comparisons with sample cross-sections show distinct differences in image characteristics based on watermelon maturity. Sugar concentration of unripe and ripe watermelon is also measured and plotted for further validation of the imaging technique.Item Novel Geospatial Data Science Techniques for Interdisciplinary Applications(University of Alabama Libraries, 2021) Sainju, Arpan Man; Jiang, Zhe; University of Alabama TuscaloosaWith the advancement of GPS and remote sensing technologies, an enormous amount of geospatial data are being collected from various domains. Examples include crime locations, road temporally detailed networks, earth observation imagery, and GPS trajectories. Geospatial data science studies computational techniques to discover interesting, previously unknown, but potentially useful patterns from large spatial datasets. It is important for various applications. Crime hotspot detection helps law enforcement departments to create effective strategies to allocate police resources and to prevent crimes. Earth observation imagery classification plays a crucial role in flood extent mapping and water resource management. Big companies like UPS use truck GPS trajectories data to find efficient routes that can ultimately minimize the delivery time and reduce carbon footprint. However, geospatial data science poses several computational challenges. First, the spatial data volume is rapidly growing. For example, NASA collects around 12TB of earth observation imagery per day. Second, spatial data exhibits spatial dependency which imply nearby samples are not statistically independent. Third, different spatial patterns of interest may exist in different spatial scales. Finally, there can be limited observations. For example, sometimes it can be difficult or even impossible to get the complete observation of spatial features in an area due the presence of obstacles (e.g., clouds). My thesis investigates novel geospatial data science techniques to address some of these challenges. I propose novel parallel spatial colocation mining algorithms on GPUs to address the challenge of large data volume. Similarly, I propose a deep learning framework to automatically map the road safety features from streetview imagery that captures spatial dependency. Next, I propose a novel approach to address the challenge of limited observation based on the physics-aware spatial structural constraint. Finally, I propose a novel spatial structured model called hidden Markov contour tree (HMCT), a contour tree structure, to capture directed spatial dependency on flow directions between all locations on a 3D surface.Item Reinforcement Learning-Based Mobile Underwater Acoustic Communications(University of Alabama Libraries, 2022) Fu, Qiang; Song, Aijun; University of Alabama TuscaloosaUnderwater acoustic communication technologies are becoming increasingly important due to the widespread adoption of autonomous unmanned vehicles (AUVs) in ocean data collection. The communication support provided by acoustic telemetry makes it possible for multiple AUVs to coordinate during underwater missions. Mobile underwater acoustic communications is still an active field of research. Various technical issues, such as reliable communications between two mobile nodes, multi-user communications, and joint optimization of navigation and communications are awaiting for satisfactory solutions. In this thesis, we develop solutions for several of these issues. The first effort considers an adaptive communication system based on time-reversed orthogonal frequency-division multiplexing methods for the underwater acoustic channels. In this adaptive system, the receiver sends truncated q-functions to the transmitter, which then performs mapping selection for the individual sub-carriers. Simulations demonstrate the advantages of the proposed adaptive system, achieving higher data rates and lower feedback costs. The second effort addresses the extended propagation delay in closed-loop adaptive communications for mobile platforms. An adaptive modulation strategy is developed based on reinforcement learning. Specifically, a Dyna-Q algorithm is presented to improve the communication throughput. Our simulations show that the Dyna-Q algorithm achieves a higher throughput and lower bit-error-rates than the direct feedback. The third effort provides a solution to the acoustic communication problem with multiple AUVs. We propose a virtual multiple-input/multiple-output (MIMO) strategy that selects transmitters from a subset of AUVs to form a virtual transmit array. A user selection algorithm is used to determine the active AUV subset for data transmissions. Adaptive modulation is combined to further improve throughput. The user selection and modulation choice are determined by the predicted data rates, thus increasing spectral efficiency of the uplink. In the last effort of this thesis, we address the trajectory optimization for underwater data muling with mobile nodes. In this scenario, multiple AUVs sample a mission area and autonomous surface vehicles (ASVs) visit underway AUVs to retrieve survey data. We propose a nearest-K reinforcement learning algorithm to optimize ASV travel tracks. The learning-based algorithm can simultaneously maximize fairness in data transmissions and minimize the travel distance of the surface nodes.Item Telerehabilitation Devices for Motion Detection and Trunk Training(University of Alabama Libraries, 2022) Zhang, Liang; Jalili, Nader; University of Alabama TuscaloosaNervous system injuries lead to various dysfunctions that needs to be restored through rehabilitation training. Not all patients are privileged to receive rehabilitation training in time. Telerehabilitation improves the accessibility of rehabilitation trainings for all patients. Yet telerehabilitation in practice is limited in the aspect of assessment and training. This dissertation documents the author's work done at UA to enable telerehabilitation. The dissertation chapters cover aspects of motion assessment, risk assessment, training independency, and device compliance. Chapter one presents a low-cost and portable walker-based human motion estimation system, which enables distance supervised walker training with objective motion and force data collected. Preliminary testing is conducted to validate the concept. Chapter two designs an RGB-camera-based fall detection algorithm in complex home environments. With low-cost components, the system maintained an accuracy of 94.5%. Chapter three presented a design of a semi-active assist as needed four-bar linkage trunk rehabilitation device. This work tested the concept of semi-active activation force, which can be implemented in future devices to increase independency of trainings. Chapter four studied flexible joints with compliance to increase user safety during rehabilitation trainings. A mechanical model of a bending spring joint actuated by a tendon is developed and simplified. With the four works contributed and future works to come, the author believes that telerehabilitation can become low cost, independent, safe, and accessible.Item Understanding microbeads stacking in deformable Nano-Sieve for Efficient plasma separation and blood cell retrieval(Elsevier, 2022) Chen, Xinye; Zhang, Shuhuan; Gan, Yu; Liu, Rui; Wang, Ruo-Qian; Du, Ke; Rochester Institute of Technology; University of Alabama Tuscaloosa; Rutgers State University New BrunswickEfficient separation of blood cells and plasma is key for numerous molecular diagnosis and therapeutics applications. Despite various microfluidics-based separation strategies having been developed, there is still a need for a simple, reliable, and multiplexing separation device that can process a large volume of blood. Here we show a microbead-packed deformable microfluidic system that can efficiently separate highly purified plasma from whole blood, as well as retrieve blocked blood cells from the device. To support and rationalize the experimental validation of the proposed device, a highly accurate model is constructed to help understand the link between the mechanical properties of the microfluidics, flow rate, and microbeads packing/leaking based on the microscope imaging and the optical coherence tomography (OCT) scanning. This deformable nano-sieve device is expected to offer a new solution for centrifuge-free diagnosis and treatment of bloodborne diseases and contribute to the design of next-generation deform able microfluidics for separation applications. (c) 2021 Elsevier Inc. All rights reserved.Item Wireless directional network: a security perspective(University of Alabama Libraries, 2019) Lu, Yu; Hu, Fei; University of Alabama TuscaloosaThe rapid growth of big data applications over wireless network such as wireless sensor networks and unman aerial vehicle (UAV) lead to higher requirements for network capacity. One of the advanced solutions is adopting multi-beam smart antenna (MBDA) technology, it enables current transmission and reception through dierent directions. However, it also introduced some new challenges for data and transmission security. In this dissertation, we would like to study the MBSA network from security perspective. Specifically, we investigated the remote data integrity, transmission secure and authentication problems in MBSA network. In Chapter 2, we will design a remote data integrity check (RDIC) scheme for graph structural data collected though MBDA network. These scheme enables verification for both data content and structural integrity and supports verifiable data update. In Chapter 3, we work on a Deep Q-Network (DQN) based defending mechanism for transmission secure in MBDA network. This defender utilizes parameters from dierent network layer to represent the system state and determine the optimal action for monitor nodes to defend the network from intentional quality of service (QoS) and node compromise. In Chapter 4, we propose a hardware testbed with USRP RIO and MBSA for directional neighbor discovery with authentication. We tested directional neighbor discovery, ID based public key authentication and directional neighbor switch based on channel quality and user priority. Also, we have a brief introduction for the background and outline of my dissertation in Chapter 1, and in Chapter 5 we summarized all works during my PhD and some possible future works I may continue.