Digital Commons Network^™

Boosted Ensemble Algorithm Strategically Trained For The Incremental Learning Of Unbalanced Data, Michael David Muhlbaier

Theses and Dissertations

Many pattern classification problems require a solution that needs to be incrementally updated over a period of time. Incremental learning problems are often complicated by the appearance of new concept classes and unbalanced cardinality in training data. The purpose of this research is to develop an algorithm capable of incrementally learning from severely unbalanced data. This work introduces three novel ensemble based algorithms derived from the incremental learning algorithm, Learn++. Learn++.NC is designed specifically for incrementally learning New Classes through dynamically adjusting the combination weights of the classifiers' decisions. Learn++.UD handles Unbalanced Data through class-conditional voting …

A "Divide-And-Conquer" Strategy For Nde Signal Inversion In Gas Transmission Pipelines, Justin Gary Bram

Theses and Dissertations

Signal inversion in nondestructive evaluation (NDE) applications is a critical step before remediation decisions are made. The accuracy and confidence of the signal inversion results therefore play a key role in evaluating the effectiveness of the NDE procedure. Conventional NDE signal inversion algorithms that employ artificial neural networks treat all geometric regions of the NDE signal equally. Consequently, when the inversion algorithm is presented with input data that is significantly different from the training data, the performance of the network deteriorates significantly. This thesis presents a superior alternative for NDE signal inversion. Different geometric regions of the NDE signature are …

Use Of Principal Component Analysis With Linear Predictive Features In Developing A Blind Snr Estimation System, Matthew James Marbach

Theses and Dissertations

Signal-to-noise ratio is an important concept in electrical communications, as it is a measurable ratio between a given transmitted signal and the inherent background noise of a transmission channel. Currently signal-to-noise ratio testing is primarily performed by using an intrusive method of comparing a corrupted signal to the original signal and giving it a score based on the comparison. However, this technique is inefficient and often impossible for practical use because it requires the original signal for comparison. A speech signal's characteristics and properties could be used to develop a non-intrusive method for determining SNR, or a method that does …

Random Feature Subspace Ensemble Based Approaches For The Analysis Of Data With Missing Features, Hussein Syed Mohammed

Theses and Dissertations

Missing data in real world applications is not an uncommon occurrence. It is not unusual for training, validation or field data to have missing features in some (or even all) of their instances, as bad sensors, failed pixels, malfunctioning equipment, unexpected noise causing signal saturation, data corruption, and so on, are all familiar scenarios in many practical applications.

In this thesis, the feasibility of an ensemble of classifiers trained on a feature subset space is investigated as an effective and practical solution for the missing feature problem. Two ensemble of classifiers approach motivated by the Random Subspace Method are proposed …

An Architecture For Intelligent Health Assessment Enabled Ieee 1451 Compliant Smart Sensors, Donald Albert Nickles

Theses and Dissertations

As systems become increasingly complex and costly, potential failure mechanisms and indicators are numerous and difficult to identify, while the cost of loss is very expensive - human lives, replacement units, and impacts to national security. In order to ensure the safety and long-term reliability of vehicles, structures, and devices attention must be directed toward the assessment and management of system health. System health is the key component that links data, information, and knowledge to action. Integrated Systems Health Management (ISHM) doctrine calls for comprehensive real-time health assessment and management of systems where the distillation of raw data into information …

Data Fusion Of Complementary Information From Parietal And Occipital Event Related Potentials For Early Diagnosis Of Alzheimer's Disease, Nicholas Stepenosky

Theses and Dissertations

The number of the elderly population affected by Alzheimer's disease is rapidly rising. The need to find an accurate, inexpensive, and non-intrusive procedure that can be made available to community healthcare providers for the early diagnosis of Alzheimer's disease is becoming an increasingly urgent public health concern. Several recent studies have looked at analyzing electroencephalogram signals through the use of many signal processing techniques. While their methods show great promise, the final outcome of these studies has been largely inconclusive. The inherent difficulty of the problem may be the cause of this outcome, but most likely it is due to …

Multiresolution Wavelet Analysis Of Event-Related Eeg Potentials Using Ensemble Of Classifier Data Fusion Techniques For Early Diagnosis Of Alzheimer's Disease, Apostolos Topalis

Theses and Dissertations

The recent advances and knowledge in medicine and nutrition have greatly improved our average life expectancy. An unfortunate consequence of this longer life span, however, is a dramatic increase in the number of individuals suffering from dementia, and more specifically, from Alzheimer's disease (AD). Furthermore, AD remains under-diagnosed and under-treated until its more severe stages due to lack of standard diagnostic tools available to community clinics. A search for biomarkers that will allow early diagnosis of the disease is therefore necessary to develop effective medical treatments. Such a biomarker should be non-invasive, simple to obtain, safe, inexpensive, accurate, and most …

Wavefront Curvature Sensing From Image Projections, Jonathan C. Buffington

Theses and Dissertations

This research outlines the development and simulation of a signal processing approach to real time wavefront curvature sensing in adaptive optics. The signal processing approach combines vectorized Charge Coupled Device (CCD) read out with a wavefront modal estimation technique. The wavefront sensing algorithm analyzes vector projections of image intensity data to provide an estimate of the wavefront phase as a combination of several low order Zernike polynomial modes. This wavefront sensor design expands on an existing idea for vector based tilt sensing by providing the ability to compensate for additional modes. Under the proposed wavefront sensing approach, the physical wavefront …

Reconstructing Spectral Scenes Using Statistical Estimation To Enhance Space Situational Awareness, Travis F. Blake

Theses and Dissertations

A new sensor, the Advanced Electro-Optical System (AEOS) Spectral Imaging Sensor (ASIS) has been developed at the Maui Space Surveillance Complex (MSSC). ASIS is capable of collecting resolved imagery of space objects in 10's-100's of spectral bands while using an adaptive optics system. However, the stringent requirements of collecting ground-based images requires a sensor that induces spectral blurring. Post-processing algorithms to remove this blurring are required to fully exploit these spectral images. This research focuses on developing the reconstruction algorithms, based on proven estimation theories, required to spectrally deblur the images collected from ASIS. Additionally, the research will expand the …

Elevated Neutral-To-Earth Voltage In Distribution Systems Including Harmonics, Jian Jiang

All Dissertations

The elevated neutral-to-earth voltage (NEV), and the related phenomenon called stray voltage, is analyzed in multigrounded distribution systems. Elevated NEV is typically caused by fundamental frequency currents returning to the source via the neutral conductor and earth. However, harmonic distortion is also found to contribute to elevated NEV. A multiphase harmonic load flow algorithm is developed to examine the effects of various factors on the NEV, including unsymmetrical system configuration, load unbalance and harmonic injection. To fulfill this objective, the system modeling is adapted to include the neutral conductor into the component equivalent circuit. The overhead transmission line is remodeled …

Haptics In Robotics And Automotive Systems, Erhun Iyasere

All Theses

Haptics is the science of applying touch (tactile) sensation and control to interaction with computer applications. The devices used to interact with computer applications are known as haptic interfaces. These devices sense some form of human movement, be it finger, head, hand or body movement and receive feedback from computer applications in form of felt sensations to the limbs or other parts of the human body. Examples of haptic interfaces range from force feedback joysticks/controllers in video game consoles to tele-operative surgery. This thesis deals with haptic interfaces involving hand movements. The first experiment involves using the end effector of …

Hierarchical State Estimation For Wide Area Power Systems, Srivatsan Lakshminarasimhan

All Theses

This thesis presents the application of hierarchical state estimation techniques to consolidate the state output of a wide area power system network. In a wide area network a large number of interconnections exist between various utilities of the wide area. Power transactions between areas occur over large distances and hence for better security there is a need to monitor the state of the entire wide area systems. Hierarchical state estimation is preferred over integrated state estimation, due to the reduced computational time.
Using existing state estimators of the member utilities of the wide area in the bottom level of hierarchical …

Nonlinear Control Techniques For Robot Manipulators, Nitendra Nath

All Theses

This Masters thesis describes the design and implementation of control strategies for the following topics of research: i) Whole Arm Grasping Control for Redundant Robot Manipulators, ii) Neural Network Grasping Controller for Continuum Robots and, iii) Coordination Control for Haptic and Teleoperator Systems.

An approach to whole arm grasping of objects using redundant robot manipulators is presented. A kinematic control which facilitates the encoding of both the end-effector position, as well as body self-motion positioning information as a desired trajectory signal for the manipulator joints is developed.

An approach is presented to whole arm grasping control for continuum robots. The …

Observation And Tracking Of Tropical Cyclones Using Resolution Enhanced Scatterometry, Richard Ryan Halterman

Theses and Dissertations

The QuikSCAT scatterometer provides global daily coverage of oceanic near-surface vector winds. Recently, algorithms have been developed to enhance the spatial resolution of QuikSCAT winds from 25~km to 2.5~km posting. These ultra-high resolution winds are used, in comparison with standard L2B data product winds, to observe and track tropical cyclones. Resolution enhanced winds are found to provide additional storm structure such as inner core size and structure and the presence of multiple eyewalls compared with standard resolution winds. The 2.5~km winds are also able to observe storms nearer to the shore than 25~km winds. An analysis of circulation center locatability …

Developing A Benchmark Suite For The Evaluation Of Orientation Sensors, Kelly Waller

All Theses

This paper examines the problem with the lack of standardization through which MEMS orientation sensors are evaluated. These sensors are sold with data sheets that outline their performance, but lack the conditions under which the testing takes place. In this research, a testing apparatus was developed, and testing routines were designed to evaluate the different characteristics of orientation sensors under different motion conditions. Three orientation sensors, each in a different price range, were evaluated with the benchmark suite. The testing apparatus is a turntable that can precisely spin an orientation sensor via a stepper motor, and can record its exact …

Channel-Access And Routing Protocols For Wireless Ad Hoc Networks With Directional Antennas, Arvind Swaminathan

All Dissertations

Medium-access control (MAC) and multiple-hop routing protocols are presented that exploit the presence of directional antennas at nodes in a wireless ad hoc network. The protocols are designed for heterogeneous networks in which an arbitrary subset use directional antennas. It is shown that the new protocols improvement the network`s performance substantially in a wide range of scenarios.
A new MAC protocol is presented that employs the RTS/CTS mechanism. It accounts for the constraints imposed by a directional antenna system, and it is designed to exploit the capabilities of a directional antenna. It is shown that the receiver blocking problem is …

A Cross-Layer Approach To Increase Spatial Reuse And Throughput For Ad Hoc Networks, Steven Boyd

All Theses

Ad hoc networks employing adaptive-transmission protocols can alter transmission parameters to suit the channel environment. Channel-access mechanisms are used to govern temporal use of the transmission medium amongst nodes. Effective operation of a channel-access mechanism can improve the ability of an adaptive-transmission protocol to accommodate changing channel conditions. The interoperability of these two mechanisms motivates cross-layer design of adaptive-transmission protocols.
In this thesis we examine the integration of a new channel-access mechanism with a physical-layer adaptive-transmission protocol to create a cross-layer protocol with enhanced capabilities. We derive specific physical-layer measurements which are used to control channel-access behavior in a distributed …

Using A Spline To Model The Motion Of A 4-Man Fireteam During Building Clearing Exercises, Marshall Werner

All Theses

This paper examines the problem of tracking positions of a fireteam (4-5 men) as they perform building clearing exercises in order to predict future motion od the team. This was done by examining the team as a single entity rather than 4 or 5 separate entities. A team model was used under the assumption that the team stays together as a group and moves in specific patterns. The model chosen for this experiment was a spline curve. This spline curve allowed for movement in straight lines through hallways as well as curving around corners.

Shape Matching, Relevance Feedback, And Indexing With Application To Spine X-Ray Image Retrieval, Xiaoqian Xu

Theses and Dissertations

The National Library of Medicine (NLM), an institute in the National Institutes of Health (NIH), maintains a collection of 17,000 digitized spine X-ray images obtained from the second National Health and Nutrition Examination Survey (NHANES II). Research effort has been devoted to develop a web-accessible retrieval system that allows retrieval of images from the NHANES II database on relevant and frequently found pathologies. A comprehensive and successful image retrieval system requires effective image representation and matching methods, relevance feedback algorithms to incorporate user opinions, and efficient indexing schemes for fast access to image databases. This dissertation studies and develops approaches …

Using Duplication With Compare For On-Line Error Detection In Fpga-Based Designs, Daniel L. Mcmurtrey

Theses and Dissertations

Space destined FPGA-based systems must employ redundancy techniques to account for the effects of upsets caused by radiated environments. Error detection techniques can be used to alert external systems to the presence of these upsets. Readback with compare is an error detection technique commonly employed in FPGA-based designs. This work introduces duplication with compare (DWC) as an automated on-line error detection technique that can be used as an alternative to readback with compare. This work also introduces a set of metrics that is used to quantify the effectiveness and coverage of this error detection technique. A tool is presented that …

Unsupervised Neural Pattern Classifiers For Oral Vowel Pronunciation Of Foreign-Accented Speech, Joseph Hecker

All Theses

This thesis describes the development of unsupervised neural-based pattern classifiers for the training of vowel pronunciations for students learning a foreign language. This paper examines American learners of the French language. A corpus of single word utterances is compiled from a group of native French speakers. Cepstral features are used to train two unsupervised neural pattern classifiers: a self-organizing map and a growing neural gas. The development and justification for the use of these classifiers is presented. The output from the classifier is translated to a bar graph for visual assessment. The degree to which the utterance sounds native is …

A High Gain Multi-Stage Operational Amplifier Using Compound Transconductance Element, Chandra Sekhar Acharyulu Durisety

Doctoral Dissertations

The rapid increase in integrated circuit complexity attributed to the advancements in fabrication techniques combined with the increasing demand for consumer applications has created an immense demand for high performance analog systems. These analog systems include high-resolution (≥ 14-bit) and/or high-speed (≥ 1 GHz) ADCs, high- linearity filters and power management circuits. The fundamental limitations in these systems are tied to the constraints imposed by the basic design elements that tend to include amplifiers and reference generators (voltage/current). With continued technology scaling, innovative circuits and design techniques are necessary in achieving high performance analog/mixed-signal systems.

The goal of this research …

A Resource Efficient Localized Recurrent Neural Network Architecture And Learning Algorithm, Daniel Borisovich Budik

Masters Theses

Recurrent neural networks (RNNs) are widely acknowledged as an effective tool that can be employed by a wide range of applications that store and process temporal sequences. The ability of RNNs to capture complex, nonlinear system dynamics has served as a driving motivation for their study. RNNs have the potential to be effectively used in modeling, system identification and adaptive control applications, to name a few, where other techniques fall short. Most of the proposed RNN learning algorithms rely on the calculation of error gradients with respect to the network weights. What distinguishes recurrent neural networks from static, or feedforward …

Robust Estimation Of Mahalanobis Distance In Hyperspectral Images, Eduardo C. Meidunas

Theses and Dissertations

This dissertation develops new estimation methods that fit Johnson distributions and generalized Pareto distributions to hyperspectral Mahalanobis distances. The Johnson distribution fit is optimized using a new method which monitors the second derivative behavior of exceedance probability to mitigate potential outlier effects. This univariate distribution is then used to derive an elliptically contoured multivariate density model for the pixel data. The generalized Pareto distribution models are optimized by a new two-pass method that estimates the tail-index parameter. This method minimizes the mean squared fitting error by correcting parameter values using data distance information from an initial pass. A unique method …

Fpga Logic Design For Analog-To-Digital-Converter Hardware Utilizing High Speed Serial Data Links, Joshua Brandon Jones

Masters Theses

The Computer Assisted Dynamic Data Monitoring and Analysis System (CADDMAS) used by the aeropropulsion test cells at Arnold Engineering Development Center (AEDC) processes a large amount of high bandwidth accelerometer and strain gage data. Data from each sensor must be digitized before being processed by software running on networked computers. This thesis describes some of the original analog-to- digital converter (ADC) hardware used by the CADDMAS as well as some of its limitations. More up-to-date ADC hardware was designed to be used in the CADDMAS to enhance capabilities required by the aeropropulsion test cells. These new capabilities included an increase …

Design And Implementation Of A Complex-Conjugate Shaper And Baseline Restorer For A Silicon-Based Neutron Detector Front-End, Jonathan Lanier Britton

Masters Theses

This thesis presents the design and implementation of a CMOS shaper with baseline restoration for use in the silicon-based neutron detector front-end to be used at the Spallation Neutron Source (SNS) at the Oak Ridge National Laboratory (ORNL). The system consists of a voltage-to-current (V-to-I) converter, a four-pole complex-conjugate semi-Gaussian current-input active filter, and a ground-sensing baseline restorer (BLR) operational transconductance amplifier (OTA). The first prototype chip Patara has been fabricated in the TSMC 0.35-micron process, and experimental results show that proper functionality was achieved. The shaper, which is influenced by a real pole prior to the V-to-I converter, has …

“Design, Development And Characterization Of A Thermal Sensor Brick System For Modular Robotics, Nikhil Arun Naik

Masters Theses

This thesis presents the work on thermal imaging sensor brick (TISB) system for modular robotics. The research demonstrates the design, development and characterization of the TISB system. The TISB system is based on the design philosophy of sensor bricks for modular robotics. In under vehicle surveillance for threat detection, which is a target application of this work we have demonstrated the advantages of the TISB system over purely vision-based systems. We have highlighted the advantages of the TISB system as an illumination invariant threat detection system for detecting hidden threat objects in the undercarriage of a car. We have compared …

A Distributed Solution For Visual Sensor Networks To Detect Targets In Crowds, Cheng Qian

Masters Theses

Visual sensor networks (VSNs), a novel concept about fulfilling vision tasks by a network of collaborative visual sensors, has been attracting more and more attentions these days. This thesis introduces some pioneering research on developing a distributed algorithm for VSNs to detect targets in a cluttered scene. The algorithm is aimed to achieve excellent performances on both detection accuracy and energy efficiency.

Based on a statistical model of the cluttered scene, the development starts with a centralized version where all the nodes send visual data to a central node and the central node invokes an iterative prioritization strategy (IPS) to …

Using Physical Compilation To Implement A System On Chip Platform, Pradeep M. Chimakurthy

Masters Theses

The goal of this thesis was to setup a complete design flow involving physical synthesis. The design chosen for this purpose was a system-on-chip (SoC) platform developed at the University of Tennessee. It involves a Leon Processor with a minimal cache configuration, an AMBA on-chip bus and an Advanced Encryption Standard module which performs decryption.

As transistor size has entered the deep submicron level, iterations involved in the design cycle have increased due to the domination of interconnect delays over cell delays. Traditionally, interconnect delay has been estimated through the use of wire-load models. However, since there is no physical …

Microprocessor Implementation Of Autoregressive Analysis Of Process Sensor Signals, Swetha Priyanka Pakala

Masters Theses

Automated signal analysis can help for effective system surveillance and also to analyze the dynamic behavior of the system such as impulse response, step response etc. Autoregressive analysis is a parametric technique widely used for system surveillance and diagnosis. The main aim objective of this research work is to develop an embedded system for autoregressive analysis of sensor signals in an online fashion for monitoring system parameters. This thesis presents the algorithm, data representation and performance of the optimized microprocessor implementation of autoregressive analysis.

In this work an autoregressive (AR) model is generated as a solution to a linear system …