Search Results (1 - 25 of 31 Results)

Sort By  
Sort Dir
 
Results per page  

Coleman, Ashley B.Feature Extraction using Dimensionality Reduction Techniques: Capturing the Human Perspective
Master of Science (MS), Wright State University, 2015, Computer Science
The purpose of this paper is to determine if any of the four commonly used dimensionality reduction techniques are reliable at extracting the same features that humans perceive as distinguishable features. The four dimensionality reduction techniques that were used in this experiment were Principal Component Analysis (PCA), Multi-Dimensional Scaling (MDS), Isomap and Kernel Principal Component Analysis (KPCA). These four techniques were applied to a dataset of images that consist of five infrared military vehicles. Out of the four techniques three out of the five resulting dimensions of PCA matched a human feature. One out of five dimensions of MDS matched a human feature. Two out of five dimensions of Isomap matched a human feature. Lastly, none of the resulting dimensions of KPCA matched any of the features that humans listed. Therefore PCA was the most reliable technique for extracting the same features as humans when given a set number of images.

Committee:

Pascal Hitzler, Ph.D. (Advisor); Mateen Rizki, Ph.D. (Committee Member); John Gallagher, Ph.D. (Committee Member)

Subjects:

Computer Science

Keywords:

Feature Extraction; Dimensionality Reduction; Principal Component Analysis; Multi-dimensional Scaling; Isomap; Kernel Principal Component Analysis

Landgraf, Andrew JGeneralized Principal Component Analysis: Dimensionality Reduction through the Projection of Natural Parameters
Doctor of Philosophy, The Ohio State University, 2015, Statistics
Principal component analysis (PCA) is very useful for a wide variety of data analysis tasks, but its implicit connection to the Gaussian distribution can be undesirable for discrete data such as binary and multi-category responses or counts. Exponential family PCA is a popular alternative to dimensionality reduction of discrete data. It is motivated as an extension of ordinary PCA by means of a matrix factorization, akin to the singular value decomposition, that maximizes the exponential family log-likelihood. We propose a new formulation of generalized PCA which extends Pearson's mean squared error optimality motivation for PCA to members of the exponential family. In contrast to the existing approach of matrix factorizations for exponential family data, our generalized PCA provides low-rank estimates of the natural parameters by projecting the saturated model parameters. Due to this difference, the number of parameters does not grow with the number of observations and the principal component scores on new data can be computed with simple matrix multiplication. When the data are binary, we derive explicit solutions of the new generalized PCA (or logistic PCA) for data matrices of special structure and provide a computationally efficient algorithm for the principal component loadings in general. We also formulate a convex relaxation of the original optimization problem, whose solution might be more effective for prediction, and derive an accelerated gradient descent algorithm. The method and algorithms for binary data are extended to other distributions, including Poisson and multinomial, and the scope of the new formulation for generalized PCA is further extended to incorporate weights, missing data, and variable normalization. These extensions enhance the utility of the proposed method for a variety of tasks such as collaborative filtering and visualization. Through simulation experiments, we compare our formulation of generalized PCA to ordinary PCA and the previous formulation to demonstrate its benefits on both binary and count datasets. In addition, two datasets are analyzed. In the binary medical diagnoses data, we show that the new logistic PCA is better able to explain and predict the probabilities than standard PCA, and is able to do so with many fewer parameters than the previous formulation. On a dataset consisting of users' song listening counts, we show that generalized PCA gives better visualization of the loadings than standard PCA and improves the prediction accuracy in a recommendation task.

Committee:

Yoonkyung Lee (Advisor); Vincent Vu (Committee Member); Yunzhang Zhu (Committee Chair)

Subjects:

Statistics

Keywords:

Binary data; Count data; Dimensionality reduction; Exponential family; Logistic PCA; Principal component analysis

Wijekoon, NishanthiSPATIAL AND TEMPORAL VARIABILITY OF SURFACE COVER IN AN ESTUARINE ECOSYSTEM FROM SATELLITE IMAGERY AND FIELD OBSERVATIONS
PHD, Kent State University, 2007, College of Arts and Sciences / Department of Geology
This study determined the capability of moderate resolution satellite imagery of 30 meter pixel dimension to investigate the spatial and temporal changes of Old Woman Creek National Estuarine Research Reserve, which is a dynamic coastal wetland of Lake Erie. Water quality and land cover reflectance data is interpreted with respect to in-situ sample measurements collected every 16 days in coincidence with the Landsat-5 TM over passing days mainly in summer 2005 and 2006. The study involved a variety of qualitative and quantitative, physical and remote sensing measurements generated from surface water and its constituents, aquatic emergent and terrestrial macrophytes, exposed mudflats, and radiometrically corrected Landsat-5 TM imagery. The prevailing environmental and climatic conditions of the area regulated the spatial and temporal variability of those land cover types.The study developed a suspended sediment concentration calibration method and two land cover variability mapping methods using Landsat-5 TM data. The two wetland mapping methods are based on principal component analysis (PCA) and scattergram segmentation of selected normalized difference remote sensing indices. In addition, the mineralogy and morphology of suspended particulates were investigated using an X-Ray diffraction (XRD) technique and environmental scanning electron microscopy (ESEM) which revealed the dominance of silica and calcite in surface water.The surface water samples provided total suspended particulate concentration (TSP) measurements which reported 0.7 correlation against normalized difference water index (NDWI) of bands 1 and 5, establishing a model to quantify TSP concentration in surface water. The principal component analysis (PCA) extracted endmember land covers reporting 87 % of total variance and their spatial and temporal distribution was mapped in order to identify the seasonal variability of macrophytes, open-water, and exposed ground. One dimensional spaces of normalized difference vegetation index (NDVI), normalized difference water index (NDWI), and normalized difference ground index (NDGI) segmented their respective scattergrams to identify the land cover interfaces in order to re-map the same land cover variability for better evaluation of the two wetland mapping techniques.

Committee:

Joseph Ortiz (Advisor)

Keywords:

Land Cover; Remote Sensing; Old Woman Creek; Wetland; Suspended Sediment; Principal Component Analysis;

Galbincea, Nicholas DCritical Analysis of Dimensionality Reduction Techniques and Statistical Microstructural Descriptors for Mesoscale Variability Quantification
Master of Science, The Ohio State University, 2017, Materials Science and Engineering
The transition of newly developed materials from the laboratory to the manufacturing floor is often hindered by the task of quantifying the material's inherit variability which spans from the atomistic to macroscale. This impedance is coupled with the task of linking this variability observed at these length scales and ultimately correlating this multidimensional variance to the macroscale performance of the material. This issue has lead to the development of statistical and mathematical frameworks for evaluating material variability. In this work, the author employs one such methodology for the purpose of mesoscale variability quantification with the goal to further explore and enhance this framework while simultaneously presenting the pathway as a computational design tool. This stochastic representation of microstructure allows for the delineation of materials to be highly dependent upon the topology of the material's structure and allows for digital representation via statistical volume elements (SVEs). Quantification of the topology of these SVEs can be achieved through utilization of statistical microstructural descriptors (SMDs), which inevitably leads to an extremely high order data set for each microstructure realization. This high order data set can then be dimensionally reduced via kernel principal component analysis (KPCA), thus allowing for the variance of the microstructure to be observed through the generation of microstructure visualizations. Enhancement of these visualizations can then be achieved through the use of the 1-way multivariate analysis of variance (1-way MANOVA). The reduced order SMD data set can then be combined with property results determined via finite element analysis (FEA) producing microstructure-property maps, thus allowing for both the microstructure and property variance to be observed graphically. Lastly, predictive models can be trained on the reduced order SMD data sets and property results utilizing the machine learning technique, neural network fitting, allowing for the generation of predictive microstructure-property surface maps. This statistical and mathematical pathway allows for the quantification of mesoscale variability and a linkage between microstructure and property variability. The overarching goal of this work will be to explore and enhance this methodology through the comparison of the linear and non-linear dimensionality techniques principal component analysis (PCA) and KPCA, in addition to exploring several SMDs effect on microstructure-property linkage and prediction, while simultaneously presenting the methodology as a computational design tool.

Committee:

Stephen Niezgoda, Dr. (Advisor); Dennis Dimiduk, Dr. (Committee Member); Soheil Soghrati, Dr. (Committee Member)

Subjects:

Materials Science; Mathematics; Statistics

Keywords:

mesoscale; microstructure variability, dimensionality reduction; statistical microstructural descriptors, material variance; two phase heterogeneous microstructure; kernel principal component analysis; machine learning; neural network fitting

Lin, MeimeiEcosystem services in a rural landscape of southwest Ohio
Master of Science, Miami University, 2012, Botany
Grasslands provide essential services and benefits to support and maintain human populations worldwide, and regulating services may be particularly important. The first chapter of this thesis reviews the breadth of regulating services provided by grasslands, including the current measurements and known approximations. Gas regulation, climate regulation, water maintenance, soil conservation, and waste treatment are the five most important regulating services. Land use/land cover change is considered as one of the most important drivers that led to the degradation of ecosystem services. The second chapter of this thesis uses multi-temporal Landsat TM image to reconstruct land use/land cover change over time, to evaluate changes in the economic values of various ecosystem services provided by each land cover class, and to determine the causes of total economic change on ecosystem services. This study provides general useful information about the gains and losses of ecosystem services due to dynamic land use/land cover change.

Committee:

M. Henry Stevens, PhD (Advisor); Mary Henry, PhD (Committee Member); Richard Moore, PhD (Committee Member)

Subjects:

Botany; Ecology; Geographic Information Science; Geography

Keywords:

Gas regulation; Carbon storage; Climate regulation; Soil retention; Water regulation; Water supply; Waste treatment; Ecosystem services; Change detection; Unsupervised classification; Principal Component Analysis

Zuzarte, Ian JerominoA Principal Component Regression Analysis for Detection of the Onset of Nocturnal Hypoglycemia in Type 1 Diabetic Patients
Master of Science in Engineering, University of Akron, 2008, Biomedical Engineering
Nocturnal hypoglycemia has been a factor in the sudden deaths of diabetic patients. Episodes of hypoglycemia in adults and children with Type 1 diabetes induce abnormalities in cardiac repolarization, including lengthening of the QT interval, QT dispersion and changes in T wave morphology. In certain circumstances, abnormally low blood glucose could be responsible for the development of a fatal cardiac arrhythmia. In this study a monitoring and alarm system was designed for detection of the onset of spontaneous nocturnal hypoglycemia through monitoring of the electrocardiogram of 24 Type I diabetic patients. It utilized the principal component regression analysis to interpret the variability of the RT interval of the ECG. It also monitored changes in the T-wave amplitude and raised alarms if abnormalities corresponding to hypoglycemia are detected. The top performance of the system was 91.60%, 100% and 85.71% for accuracy, sensitivity and specificity respectively, which were statistically comparable to those obtained for the system using the tangent method. This study supports the proposition that a relationship exists between cardiac function and abnormally low blood glucose.

Committee:

Dale Mugler, PhD (Advisor)

Subjects:

Biomedical Research; Engineering

Keywords:

Nocturnal hypoglycemia; Type 1 diabetes; Prolonged QTc; T-wave; Principal Component Analysis; Principal Component Regression

Raiford, Douglas WhitmoreMultivariate Analysis of Prokaryotic Amino Acid Usage Bias: A Computational Method for Understanding Protein Building Block Selection in Primitive Organisms
Master of Science (MS), Wright State University, 2005, Computer Science
Raiford III, Douglas . M.S., Department of Computer Science & Engineering, Wright State University, 2005. Multivariate analysis of prokaryotic amino acid usage bias: a computational method for understanding protein building block selection in primitive organisms. Organisms expend a significant fraction of their overall energy budget in the creation of proteins, particularly for those that are produced in large quantities. Recent research has demonstrated that genes encoding these proteins are shaped by natural selection to produce the proteins with low cost building blocks (amino acids) whenever possible. The negative correlation between protein production rate and their energetic costs has been established for two bacterial genomes: Escherichia coli and Bacillus subtilis. This thesis provides scientific validation of this theory by automating the analysis and extending the research to additional genomes. Investigations into building block selection are highly computational in nature. Diverse methodologies, including principal component analysis, calculation of Mahalanobis distance, and the execution of Mantel-Haenszel and Bonferroni tests, are required in order to automate the process. In order to verify that the cause of the observed trend is energetic cost minimization it is necessary to eliminate as many alternative explanations as possible. This is accomplished through demonstration that the trend is not localized to any particular region of the protein’s primary structure and that the trend is consistent across all genes regardless of functionality. This investigation of the energetic cost of polypeptide synthesis provides valuable insights into protein building block selection. As an example, parasitic organisms appear to exhibit no correlation between protein production rate and amino acid cost. When the costs associated with building blocks that the parasite obtains from its host are removed,however, a trend once again becomes evident.

Committee:

Michael Raymer (Advisor)

Subjects:

Computer Science

Keywords:

metabolic efficiency; codon usage bias; amino acid; PCA; principal component analysis; protein synthesis; expressivity

Woodward, Stephen M.PRINCIPAL COMPONENT ANALYSIS OF SEDIMENT DEPOSITED IN THE VILLAGE OF TITIANA FROM THE SOLOMON ISLANDS TSUNAMI OF APRIL 2, 2007
MS, Kent State University, 2009, College of Arts and Sciences / Department of Geology
On April 2 2007, an earthquake of M8.0 struck the Solomon Islands. The earthquake caused a large tsunami, resulting in over 40 deaths. Understanding the flow characteristics, destructive potential and recurrence interval for tsunamis such as this one enable future events to be anticipated and adequately planned for in terms of loss of human life, destruction of infrastructure and economic impacts. In areas like the Solomon Islands, which posses a relatively short written history, this kind of information can only be obtained through the geological record, which must be interpreted by first looking at present events and applying them to past events (Jaffe and Gelfenbaum, 2002). Most seminal tsunami recurrence and grain size studies, such as Atwater and Moore (1992), have been done in temperate siliciclastic settings. Flow characteristics inferred from carbonate sediment deposits are much harder to interpret because the shape of the grain is much more important than the size (Maiklem, 1968) (Braithwaite, 1973), making a traditional study of landward fining trends less significant. This study takes the field data I gathered in the Solomon Islands and uses principal component analysis of the sediment data to determine where the sediment is being transported from and say something about how carbonate sediment populations move and interact during a tsunami.

Committee:

Joseph Ortiz, Dr. (Committee Chair); Andrew Moore, Dr. (Advisor); Neil Wells, Dr. (Advisor)

Subjects:

Experiments; Fluid Dynamics; Geology; Soil Sciences

Keywords:

tsunami; carbonate; Principal Component Analysis

Mohiddin, Syed BDevelopment of novel unsupervised and supervised informatics methods for drug discovery applications
Doctor of Philosophy, The Ohio State University, 2006, Chemical Engineering
As of 2002, the cost of discovering a new drug was nearly $802 million with a timeline of nearly 13.6 years. Despite the large investments in time and money, drugs that were successfully introduced in the market had to be withdrawn later due to efficacy (38%) and safety (20%) reasons. Improving the success rate in drug discovery is linked with two key steps in the process. First, in order to improve efficacy, there is a need for improved understanding of genetic biomarkers (targets for drug action) that are responsible for characterizing a given disease. Second, it is possible to improve drug safety, by predicting the activity/toxicity of potential drug candidates at an early stage prior to the initiation of expensive clinical trials. In this work, we develop a novel unsupervised informatics methodology that addresses characterization of both biological and chemical samples and identification of underlying key non-redundant features responsible for characterization. Biological samples are characterized into different groups (e.g. cancer types) based on gene expression profiling and the genetic biomarkers most responsible for characterization are identified. Similarly, chemical compounds are characterized into different groups with varying activity/toxicity based on structural, physical and chemical property data of the chemical compounds. The methodology developed in this work relies largely on the multivariate aspects of principal component analysis and the application of k-means clustering algorithm in a hierarchically recursive manner to achieve unsupervised multi-class classification. The principal components are replaced by the corresponding partial least square (PLS) components in the supervised scenario. Selection of influential components (principal components in unsupervised case and PLS components in supervised case) for the purpose of classification is demonstrated and is one of the key steps for the success of this methodology. Hierarchical k-means is applied recursively to achieve binary classification at each stage eventually resulting in multi-class classification. Identification of features responsible for classification is achieved by examining the appropriate loadings of the principal or PLS components along with their coefficient of correlation with influential components.

Committee:

James Rathman (Advisor)

Subjects:

Engineering, Chemical

Keywords:

Unsupervised Classification; Supervised Classification; Principal Component Analysis; Partial Least Squares; Hierarchical K-means Clustering; Identifying Diverse Molecular Targets; Prediction of Toxicity/Activity of Chemicals

Brown, Michael J.SINGULAR VALUE DECOMPOSITION AND 2D PRINCIPAL COMPONENT ANALYSIS OF IRIS-BIOMETRICS FOR AUTOMATIC HUMAN IDENTIFICATION
Master of Science (MS), Ohio University, 2006, Electrical Engineering & Computer Science (Engineering and Technology)

With the recent emphasis given to security, automatic human identification has received significant attention. In particular, iris based subject recognition has become especially important because of its high level of complexity which lends itself to high confidence recognition. In addition, the eye is well protected and generally does not change very much over extended periods of time. This thesis gives a review of some currently available methods that have already been investigated. A wide sense stationary approximation for gray scale values is explored as a possible means of feature extraction. The singular value decomposition (SVD) is discussed as a low bit rate tool for iris discrimination. The 2D principal component analysis (2DPCA) is explored as a method for feature extraction. It is determined experimentally that the SVD for iris recognition is a novel way to significantly reduce the storage requirements (133 bits) for iris recognition as compared to other methods (2048 bits). However, recognition accuracy has not reached a desirable level. The 2DPCA, on the other hand, significantly improves recognition accuracy on the same dataset, but at the cost of greater storage requirements.

Committee:

Mehmet Celenk (Advisor)

Keywords:

iris recognition; singular value decomposition (SVD); 2D principal component analysis (2DPCA); biometrics

Orton, Christopher RobertThe Multi-Isotope Process Monitor: Non-destructive, Near-Real-Time Nuclear Safeguards Monitoring at a Reprocessing Facility
Doctor of Philosophy, The Ohio State University, 2009, Nuclear Engineering
The IAEA will require advanced technologies to effectively safeguard nuclear material at envisioned large scale nuclear reprocessing plants. This dissertation describes results from simulations and experiments designed to test the Multi-Isotope Process (MIP) Monitor, a novel safeguards approach for process monitoring in reprocessing plants. The MIP Monitor combines the detection of intrinsic gamma ray signatures emitted from process solutions with multivariate analysis to detect off-normal conditions in process streams, nondestructively and in near-real time (NRT). Three different models were used to predict spent nuclear fuel composition, estimate chemical distribution during separation, and simulate spectra from a variety of gamma detectors in product and raffinate streams for processed fuel. This was done for fuel with various irradiation histories and under a variety of plant operating conditions. Experiments were performed to validate the results from the model. Three segments of commercial spent nuclear fuel with variations in burnup and cooling time were dissolved and subjected to a batch PUREX method to separate the uranium and plutonium from fission and activation products. Gamma spectra were recorded by high purity germanium (HPGe) and cadmium zinc telluride (CZT) detectors. Hierarchal Cluster Analysis (HCA) and Principal Component Analysis (PCA) were applied to spectra from both model and experiment to investigate spectral variations as a function of acid concentration, burnup level and cooling time. Partial Least Squares was utilized to extract quantitative information about process variables, such as acid concentration or burnup. The MIP Monitor was found to be sensitive to the induced variations of the process and was capable of extracting quantitative process information from the analyzed spectra.

Committee:

Richard Christensen, PhD (Advisor); Richard Denning, PhD (Committee Member); Xiaodong Sun, PhD (Committee Member)

Subjects:

Engineering; Nuclear Chemistry; Radiation

Keywords:

nuclear safeguards; gamma spectroscopy; principal component analysis; reprocessing

Aradhye, Hrishikesh BalkrishnaAnomaly Detection Using Multiscale Methods
Doctor of Philosophy, The Ohio State University, 2001, Chemical Engineering
In an environment where most process maneuvers are automated, algorithms to detect and classify abnormal trends in process measurements are of critical importance. The petrochemical industry in the United States loses billions of dollars annually due to improper abnormal situation management, and a staggering one in 16 plant accidents results in a fatality. Hence, Statistical Process Control and Monitoring (SPC) has been an active area of research for many decades and a variety of statistical and machine learning-based methods have been developed. However, most existing methods for process monitoring learn the signal characteristics at a fixed scale, and are best for detecting changes at that single scale. In contrast, data from most industrial processes are inherently multiscale in nature due to events occurring with different localization in time, space, and frequency. Unfortunately, existing techniques are unable to adapt automatically to the scale of these features. Many existing methods also require the measurements to be uncorrelated, whereas, in practice, autocorrelated measurements are very common in industrial processes. In this work, we have investigated the use of multiscale techniques to improve upon these shortcomings of existing single-scale approaches. Because of fundamental functional relationships such as process chemistry, energy and mass balances, measurements in multivariate processes are correlated. Our approach learns these correlations and clustering behaviors in the wavelet space using machine learning methods such as Adaptive Resonance Theory (ART-2) and Principal Component Analysis (PCA), resulting in higher detection accuracy coupled with noise reduction. The performance of our method, named Multi-Scale Statistical Process Control and Monitoring (MSSPC), is compared with existing methods based on the average detection delays for detecting shifts of different sizes. Our ART-2 based MSSPC detector is currently deployed in a large scale petrochemical plant to detect process anomalies in real time by incrementally learning normal process operation in the wavelet domain. Several case studies for the detection of real process malfunctions, including the comparison with the performance of human operators, are also presented in this work. These results indicate that MSSPC is a good method for monitoring of measurements with unknown and different types of changes.

Committee:

James Davis (Advisor)

Subjects:

Engineering, Chemical

Keywords:

Wavelets; Principal Component Analysis; Adaptive Resonance Theory; Statistical Process Monitoring; Change Detection

Chivers, Daniel StephenHuman Action Recognition by Principal Component Analysis of Motion Curves
Doctor of Philosophy (PhD), Wright State University, 2012, Computer Science and Engineering PhD

Human action recognition is used to automatically detect and recognize actions per- formed by humans in a video. Applications include visual surveillance, human-computer interaction, and robot intelligence, to name a few. An example of a surveillance application is a system that monitors a large public area, such as an airport, for suspicious activity. In human-machine interaction, computers may be controlled by simple human actions. For example, the motion of an arm may instruct the computer to rotate a 3-D model that is being displayed. Human action recognition is also an important capability of intelligent robots that interact with humans.

General approaches to human action recognition fall under two categories: those that are based on tracking and those that do not use tracking. Approaches that do not use tracking often cannot recognize complex motions where movement of different parts of the body is important. Tracking-based approaches that use motion of different parts of the body are generally more powerful but are computationally more expensive, making them inappropriate for applications that require real-time responses.

We propose a new approach to human action recognition that is able to learn various human actions and later recognize them in an efficient manner. In this approach, motion trajectories are formed by tracking one or more key points on the human body. In particular, points on the hands and feet are tracked. A curve is fitted to each motion trajectory to smooth noise and to form a continuous and differentiable curve. A motion curve is then segmented into “basic motion” segments by detecting peak curvature points. To recognize an observed basic motion, a vector of curve features describing the motion is created, the vector is projected to the eigenspace created during PCA training, and the action most similar to a learned action is identified using the k-nearest neighbor decision rule.

The proposed approach simplifies action recognition by requiring that only a small number of points on a subject's body be tracked. It is shown that the motion curves obtained by tracking a small number of points are sufficient to recognize various human actions with a high degree of accuracy.

Furthermore, the proposed approach can improve the recognition power of other ap- proaches by recognizing detailed basic motions, such as foot steps, while introducing ef- ficient tracking and recognition compared to previous approaches. Recognition of basic motions allows a high-level recognizer to recognize more complex or composite actions by using the proposed system as a low-level recognizer.

Contributions of this work include reducing each video frame to a few key points on the subject's body, using curve fitting to smooth trajectory data and provide reliable seg- mentation of the motion, and efficient recognition of basic motions using PCA.

Committee:

Arthur Goshtasby, PhD (Advisor); Mateen Rizki, PhD (Committee Member); Thomas Wischgoll, PhD (Committee Member); David Miller, PhD (Committee Member)

Subjects:

Computer Engineering; Computer Science

Keywords:

Human Action Recognition; Motion Analysis; Principal Component Analysis

Chung, Koon Yin C.Facial Expression Recognition by Using Class Mean Gabor Responses with Kernel Principal Component Analysis
Master of Science (MS), Ohio University, 2010, Computer Science (Engineering and Technology)
This thesis presents a novel approach for recognizing facial expressions by incorporating class-mean Gabor responses of sampled images of human facial expressions and kernel principal component analysis (kernel PCA) with fractional polynomial power models. A mean vector of features is obtained with Gabor filters from a class of images instead of the more common method in which features are obtained from individual images. The computational cost of spatial convolutions on mean features of a class is less than the same type of convolutions with individual features. The dimensionality of mean features from Gabor filters is further reduced by using a kernel PCA technique with polynomial kernels. The kernel PCA technique is extended to use fractional power polynomial models for facial expression recognition. The proposed approach has the advantage of doing fewer projections than other facial expression recognition approaches that use traditional kernel PCA models. The proposed approach of class-mean Gabor responses has higher accuracy than existing systems that use the kernel PCA technique with class-mean image responses only.

Committee:

David M. Chelberg, PhD (Advisor); Jun Dong Liu, PhD (Committee Member); Frank Drews, PhD (Committee Member)

Subjects:

Computer Science; Electrical Engineering; Engineering

Keywords:

Facial Expression Recognition; Gabor; Kernel Principal Component Analysis; KPCA; Class Mean Gabor Responses

Ghosh Dastidar, SamanwoyModels of EEG data mining and classification in temporal lobe epilepsy: wavelet-chaos-neural network methodology and spiking neural networks
Doctor of Philosophy, The Ohio State University, 2007, Biomedical Engineering
A multi-paradigm approach integrating three novel computational paradigms: wavelet transforms, chaos theory, and artificial neural networks is developed for EEG-based epilepsy diagnosis and seizure detection. This research challenges the assumption that the EEG represents the dynamics of the entire brain as a unified system. It is postulated that the sub-bands yield more accurate information about constituent neuronal activities underlying the EEG. Consequently, certain changes in EEGs not evident in the original full-spectrum EEG may be amplified when each sub-band is analyzed separately. A novel wavelet-chaos methodology is presented for analysis of EEGs and delta, theta, alpha, beta, and gamma sub-bands of EEGs for detection of seizure and epilepsy. The methodology is applied to three different groups of EEGs: healthy subjects, epileptic subjects during a seizure-free interval (interictal), and epileptic subjects during a seizure (ictal). Two potential markers of abnormality quantifying the non-linear chaotic EEG dynamics are discovered: the correlation dimension and largest Lyapunov exponent. A novel wavelet-chaos-neural network methodology is developed for EEG classification. Along with the aforementioned two parameters, the standard deviation (quantifying the signal variance) is employed for EEG representation. It was discovered that a particular mixed-band feature space consisting of nine parameters and LMBPNN result in the highest classification accuracy (96.7%). To increase the robustness of classification, a novel principal component analysis-enhanced cosine radial basis function neural network classifier is developed. The rearrangement of the input space along the principal components of the data improves the classification accuracy of the cosine radial basis function neural network employed in the second stage significantly. The new classifier is as accurate as LMBPNN and is twice as robust. Next, biologically realistic artificial neural networks are developed to reach the next milestone in artificial intelligence. First, an efficient spiking neural network (SNN) model is presented using three training algorithms: SpikeProp, QuickProp, and RProp. Three measures of performance are investigated: number of convergence epochs, computational efficiency, and classification accuracy. Next, a new Multi-Spiking Neural Network (MuSpiNN) and supervised learning algorithm (Multi-SpikeProp) are developed. Finally, the models are applied to the epilepsy and seizure detection problems to achieve high classification accuracies.

Committee:

Hojjat Adeli (Advisor)

Keywords:

Temporal Lobe Epilepsy; Electroencephalogram (EEG); EEG Classification; Epilepsy Diagnosis; Seizure Detection; Wavelet Transform; Chaos Theory; Artificial Neural Networks; Spiking Neural Networks; Principal Component Analysis; Cosine Radial Basis Function

Yengwia , Lawrenzo NAnalyzing Recycling Habits in Mahoning County, Ohio
Master of Science in Environmental Science, Youngstown State University, 2017, Department of Geological & Environmental Sciences
Solid waste management is a challenge for municipal locality and a global issue as population and consumption growth results in increasing the quantities of waste. The amount of recyclables generated commences with personal consumption pattern and habits. In addition, the success of recycling programs such as curbside collection and drop off recycling site greatly depends on the household participation as well as community contribution. Demographic data was obtained from the US Census Bureau while the recycling outreach program and recycling data for a five year period (2008-2012) was obtained from the Mahoning County Waste Management District also known as the Green Team. The Green Team is responsible to manage the waste in Mahoning County. Principal component analysis (PCA), Analysis of Covariance (ANCOVA) was used to evaluate the relationship between demographic factors, recycling outreach programs and the amount of recycling per capita. And a GIS descriptive analysis illustrated the visual analysis of the PCA results. Demographic factors such as; population, population density, income, age, household size occupancy, were used to evaluate the amount of recycling per capita. Principal component analysis (PCA) performed on the demographic factors created 21components but only 7 components were considered important for the analysis based on their individual eigenvalue above 1. The scores of the first 2 PCA components (PCA1 and PCA2) containing about 50 percent of all the variables were plotted against the amount of recycling per capita. Then the result was illustrated on a biplot with bubble size represented the amount of recycling and bubbles with similar demographic factors clustered together. An ANCOVA analysis was performed using the following variables, PCA 1 and PCA 2 scores, School outreach programs, adult outreach program, and average recycling for five years. A P-value of 0.048 indicated a correlation between the adult school program and the amount of recycling. Pearson correlation to statistically test did not illustrate any significant correlation between similar factors as the ANCOVA. GIS descriptive analysis conducted described the different distribution and also shows township with high and low amount of recycling. In conclusion, PCA analysis showed combinations of factors such as income, educational attainment, household size occupancy, were related to the amount of recycling per capita.

Committee:

Feliciajavascript:addAdvisor(); Armstrong, PhD (Advisor); Peter Kimosop, PhD (Committee Member); Tony Vercellino, PhD (Committee Member); Daniel Kuzma (Committee Member)

Subjects:

Demographics; Environmental Science; Geographic Information Science; Statistics

Keywords:

Recycling habits in Mahoning County, Ohio;using demographic factors;principal component analysis PCA;green team recycling outreach programs

Coleman, Jill S. M.Atmospheric circulation types associated with cause-specific daily mortality in the central United States
Doctor of Philosophy, The Ohio State University, 2005, Geography
The relationship between weather and human mortality has been limited to studies centering on one or two meteorological variables. Recently, the focus is shifting toward a more holistic approach to weather, with several meteorological elements being combined to describe an air mass or weather situation that can effect the human environment. This dissertation continues the trend of recognizing discrete atmospheric circulation types that are most associated with negative health effects and identifies those causes of death occurring most regularly with a specific weather situation. Seventeen MSAs are investigated for significant mortality rate differences among a set of synoptic types prevalent in the central United States, a region that encounters frequent changes in weather conditions. Atmospheric circulation types are determined using statistical techniques. S-mode principal component analysis (PCA) is performed on a 7,305 x 1,225 ((the number of days between 1979-1998) x (25 atmospheric variables x 49 grid points)) matrix. The unrotated score solution is then entered into a two-step clustering procedure to resolve days into groups with similar atmospheric circulation conditions. The final cluster solution categorizes each day into one of nine synoptic types that characterize seasonal changes in atmospheric circulation over the central and eastern United States. Three of the synoptic types occur dominantly in summer and the winter types show some interannual variation with modes of Pacific atmospheric teleconnections. The nine synoptic classes are compared using the ANOVA F-test statistic for significant differences in daily mortality for each MSA by mortality grouping (e.g., respiratory deaths) and season. Significant differences in mortality rates among the synoptic types are most prevalent during spring and autumn at a one-day lag period, especially between the generally lower warm-season cluster mortality rates and higher rates in cold season clusters. Mortality rates are significantly different among the synoptic types for people over age 65, for weather related causes, and for circulatory death categories. Significant results for winter and especially summer are more irregular and do not display the sharp contrast between the warm season and cooler-colder season cluster sets.

Committee:

Jeffrey Rogers (Advisor)

Subjects:

Physical Geography

Keywords:

weather type; mortality; principal component analysis; cluster analysis

Aljarrah, Inad AThree Dimensional Face Recognition Using Two Dimensional Principal Component Analysis
Doctor of Philosophy (PhD), Ohio University, 2006, Electrical Engineering & Computer Science (Engineering and Technology)

This dissertation describes a face recognition system that overcomes the problem of changes in gesture and mimics in three-dimensional (3D) range images. Here, we propose a local variation detection and restoration method based on the two-dimensional (2D) principal component analysis (PCA). The depth map of a 3D facial image is first thresholded to discard the background information. The detected face shape is normalized to a standard image size of 100x100 pixels and the forefront nose point is selected to be the image center. Facial depth-values are scaled between 0 and 255 for translation and scaling-invariant identification. The preprocessed face image is smoothed to minimize the local variations. The 2DPCA is applied to the resultant range data and the corresponding principal- (or eigen-) images are used as the characteristic feature vectors of the subject to find his/her identity in the database of pre-recorded faces.

The system’s performance is tested against the GavabDB and Notre Dame University facial databases. Experimental results show that the proposed method is able to identify subjects with different gesture and mimics in the presence of noise in their 3D facial images.

Committee:

Mehmet Celenk (Advisor)

Keywords:

3D face recognition; 2D principal component analysis; computer vision

FANEGAN, JULIUS BOLUDEA FUZZY MODEL FOR ESTIMATING REMAINING LIFETIME OF A DIESEL ENGINE
MS, University of Cincinnati, 2007, Engineering : Electrical Engineering
The research is a novel fuzzy modeling approach built on Sugeno and Yasukawa’s qualitative modeling algorithm; however, it differs in the identification step. More precisely, the current approach applies dimensionality reduction techniques in the identification step. The approach is illustrated on the problem of estimating the remaining lifetime of a diesel engine and it is applied to a real data set. This data set represents a total of nine diesel engine trucks, and for each engine, it consists of 13 sensors recorded in time.

Committee:

Dr. Anca Ralescu (Advisor)

Keywords:

Principal Component Analysis (PCA); Fuzzy Modeling; C-Means Clustering

Cui, ChenAdaptive weighted local textural features for illumination, expression and occlusion invariant face recognition
Master of Science (M.S.), University of Dayton, 2013, Electrical Engineering
Face recognition is one of the most promising biometric methodologies for human identification in a non-cooperative security environment. Several algorithms have been developed for extracting different facial features for face recognition. Due to the various possible challenges of data captured at different lighting conditions, viewing angles, facial expressions, and partial occlusions in natural environmental conditions, automatic facial recognition is still remaining as a difficult issue that needs to be resolved. In this thesis, we propose a novel approach to tackle some of these issues by analyzing the local textural descriptions for facial feature representation. The textural information is extracted by an enhanced local binary pattern description of all the local regions in the face, and the dimensionality of the texture image is reduced by principal component analysis performed on each local face region independently. The face feature vector is obtained by concatenating the reduced dimensional weight set of each module (sub-region) of the face image. The weightage of each sub-region is determined by employing the local variance estimate of the respective region which represents the significance of the region. Experiments conducted on various popular face databases show promising performance of the proposed algorithm in varying lighting, expression, and partial occlusion conditions. Research work is progressing to investigate the effectiveness of the proposed face recognition method on pose varying conditions as well. It is envisaged that a multilane approach of trained frameworks at different pose bins and an appropriate voting strategy would lead to a good recognition rate in such situation.

Committee:

Vijayan Asari, Ph.D. (Committee Chair); Balster Eric, Ph.D. (Committee Member); Ordóñez Raúl, Ph.D. (Committee Member)

Subjects:

Electrical Engineering

Keywords:

Enhanced Local Binary Pattern; Weighted Modular Principal Component Analysis; Face Recognition; Feature Extraction

Ergin, Leanna NENHANCED DATA REDUCTION, SEGMENTATION, AND SPATIAL MULTIPLEXING METHODS FOR HYPERSPECTRAL IMAGING
Doctor of Philosophy in Clinical-Bioanalytical Chemistry, Cleveland State University, 2017, College of Sciences and Health Professions
A hyperspectral image is a dataset consisting of both spectra and spatial information. It can be thought of either as a full spectrum taken at many pixel locations on a sample or many images of the same sample, each at a different wavelength. In recent decades, hyperspectral imaging has become a routine analytical method due to rapid advances in instrumentation and technique. Advances such as the speed of data acquisition, improved signal-to-noise-ratio, improved spatial resolution, and miniaturization of the instrumentation have all occurred, making chemical imaging methods more robust and more widely used. The work presented here deals with three issues in the field of hyperspectral imaging: unassisted data processing that is chemically meaningful and allows for subsequent chemometric analyses, visualization of the data that utilizes the full colorspace of modern red, green, blue (RGB) displays, and data collection with improved signal-to-noise ratios and comparably short acquisition times. Hyperspectral image data processing is a fundamental challenge in the field. There is a need for reliable processing techniques that can operate on the large amount of data in a hyperspectral image dataset. Because of the large quantity of data, currently-used methods for data processing are problematic because of how time-consuming and calculation-intensive they are or because of increased error that is observed in the less-intensive methods. The work presented here includes a user-unassisted method for rapidly generating chemical-based image contrast from hyperspectral image data. Our method, reduction of spectral images (ROSI), is an effective hyperspectral image processing method. A full theoretical description of the method is given along with performance metrics. The description has been generalized to work with any number of wavelength dimensions and spectra. A concise protocol is put forth that will enable other researchers to utilize this method by following a short, simple list of steps. ROSI can also be used as a data reduction method, as it achieves a threshold information density in the spectral dimension for all image pixels. ROSI results are suitable for subsequent data analysis enabling ROSI to be performed alone or as a preprocessing data reduction step. This research also improves upon a spatially-multiplexed Raman imaging system based on the digital micromirror device (DMD). The system provides signal-to-noise ratio enhancement while maintaining laser powers below the damage threshold of the sample and comparably short acquisition times. In the work presented here, the spatial resolution of the DMD imager has been improved such that features with a width of 2.19µm could be resolved, whereas the previous limit was 7.81µm.

Committee:

John Turner, II, Ph.D. (Advisor); David Ball, Ph.D. (Committee Member); Petru Fodor, Ph.D. (Committee Member); Xue-Long Sun, Ph.D. (Committee Member); Yan Xu, Ph.D. (Committee Member); Aimin Zhou, Ph.D. (Committee Member)

Subjects:

Analytical Chemistry; Chemistry; Scientific Imaging

Keywords:

hyperspectral imaging; chemical imaging; data reduction; image segmentation; direction cosine; principal component analysis; PCA; reduction of spectral angle; ROSI; digital micromirror device; spatial multiplexing;

Henry, Emily BrookeStochastic Modeling of Geometric Mistuning and Application to Fleet Response Prediction
Master of Science in Engineering (MSEgr), Wright State University, 2014, Mechanical Engineering
An improved spatial statistical approach and probabilistic prediction method for mistuned integrally bladed rotors is proposed and validated with a large population of rotors. Prior work utilized blade-alone principal component analysis to model spatial variation arising from geometric deviations contributing to forced response mistuning amplification. Often, these studies considered a single rotor measured by contact probe coordinate measurement machines to assess the predictive capabilities of spatial statistics through principal component analysis. The validity of the approach has not yet been demonstrated on a large population of mistuned rotors representative of operating fleets, a shortcoming addressed in this work. Furthermore, this work improves the existing predictions by applying principal component methods to sets of airfoil (rotor) measurements, thus effectively capturing blade-to-blade spatial correlations. In conjunction with bootstrap sampling, the method is validated with a set of 40 rotors and quantifies the subset size needed to characterize the population. The work combines a novel statistical representation of rotor geometric mistuning with that of probabilistic techniques to predict the known distribution of forced response amplitudes.

Committee:

Joseph C. Slater, Ph.D., P.E. (Advisor); Jeffrey M. Brown, Ph.D. (Committee Member); J. Mitch Wolff, Ph.D. (Committee Member); Ha-Rok Bae, Ph.D. (Committee Member)

Subjects:

Aerospace Engineering; Engineering; Mechanical Engineering

Keywords:

Principal Component Analysis; PCA; Mistuning; Geometric Mistuning; Probabilistics; Bootstrapping; Turbine Engine; Rotor; Integrally Bladed Rotor; IBR; Blisk; Forced Response; Prediction

Kaufman, Jason R.Digital video watermarking using singular value decomposition and two-dimensional principal component analysis
Master of Science (MS), Ohio University, 2006, Electrical Engineering & Computer Science (Engineering and Technology)

As the state of remote sensing technology improves, the acquisition of three-dimensional images and video will become more common in several different applications. However, the problem of protecting and authenticating three-dimensional data – in particular, three-dimensional video data – has been largely unexplored. An application of the singular value decomposition (SVD) and two-dimensional principal component analysis (2DPCA) to video data with an arbitrary number of channels for the purpose of watermarking is presented.

It will be shown that it is possible to select parameters that preserve the visual quality of the video while effectively embedding the watermark in both the spatial and temporal domains. However, much processing time is required to embed and extract the watermark. Furthermore, it is unclear how robust the presented technique is to attack.

Committee:

Mehmet Celenk (Advisor)

Keywords:

digital video watermarking; information security; singular value decomposition (SVD); two-dimensional principal component analysis (2DPCA)

Bonini, NickAssessing the Variability of Phytoplankton Assemblages in Old Woman Creek, Ohio
PHD, Kent State University, 2016, College of Arts and Sciences / Department of Geology
Various techniques for assessing, monitoring, and predicting algal blooms in an estuarine ecosystem are analyzed. In one section, routine water samples are collected at previously established monitoring sites in Old Woman Creek, filtered onto a 47 mm, 0.7 µm glass-fiber filter (GF/F), and then measured using a visible/near-infrared spectrophotometer. Varimax-rotated principal component analysis (VPCA) is applied to reflectance data and then used to quantify and identify pigments, phytoplankton taxa, and sediments by comparing the measured spectral signatures to known standards. Common assemblages that are reported throughout the three-year study include: bacillariophyceae (diatoms), chlorophyta (green algae), cyanobacteria (blue-green algae), and illite. A similar approach is taken in the next section by applying multivariate statistics to Landsat 8 satellite imagery in order to determine the distribution of in-water constituents at a high spatial resolution. Only four bands in the visible range are available for this analysis, but it is possible to identify several of the same groups of algae and sediments, providing a useful complement to the hyperspectral work. Finally, a bloom prediction model based on springtime discharge is created by applying VPCA to in-water sonde data from one of the monitoring sites at Old Woman Creek during a recent 11-year time period. In this model, a proxy for net community production (NCP) is determined using oxygen and pH dynamics and then compared to daily rates of streamflow. Possible monthly sequences between January and June are considered in order to determine which timeframe is the best indicator of the average annual NCP. Time of day (daytime versus nighttime) and mouth bar conditions (barrier beach present versus absent) are important factors in determining production in the estuary. Based on the results, the best predictor for NCP is stream discharge from March through May, which produces correlations that are significant at even the 1% level. A positive relationship is found between NCP and discharge when flow from Old Woman Creek into Lake Erie is permitted. When flow is blocked by the barrier beach, however, the relationship is reversed.

Committee:

Joseph Ortiz (Advisor); Anne Jefferson (Committee Member); Alison Smith (Committee Member); Darren Bade (Committee Member)

Subjects:

Aquatic Sciences; Biological Oceanography; Environmental Geology; Environmental Science; Geology; Limnology; Water Resource Management

Keywords:

Old Woman Creek; Lake Erie; algal blooms; phytoplankton; water quality; estuary; barrier beach; VNIR derivative spectroscopy; VPCA; principal component analysis; remote sensing; prediction model; net community production; streamflow

Hong, SoonyoungAn effective data mining approach for structure damage indentification
Doctor of Philosophy, The Ohio State University, 2007, Aeronautical and Astronautical Engineering
An efficient, neural network based, online nondestructive structural damage identification procedure is developed for determining the damage characteristics (the damage locations and the corresponding severity) from dynamic measurements in near real-time. The procedure utilizes unique data processing techniques to track the most useful modal information based on modal strain energy and to calculate the associated data based on principal component analysis for further processing in a neural network based identification scheme. With two unique features, this approach is significantly different from currently available damage identification procedures for real-time structural integrity monitoring/diagnostics. First, the most sensitive mode for the specific damage is selected in an automatic process which increases the accuracy of damage identification and decreases time spent on neural network training. Second, the approach creates unique data that extracts core characteristics from modal information for a number of different damage cases; and consequently, the accuracy of the damage identification improves significantly. This approach can be operated online providing real time structural damage identification. The method is tested for simulated damage cases, including situations of single and multiple damage in the closely-spaced frequencies of Kabe's model. The philosophy behind the proposed research is to provide a means to online and nondestructively predict the degradation of a structure's integrity (i.e. damage location and the corresponding severity, strength loss).

Committee:

Mo-How Shen (Advisor)

Keywords:

Structure Health Monitoring; Vibration Based Damage Identification; Principal Component Analysis; Modal Strain Energy

Next Page