Doctoral Thesis
Refine
Year of publication
- 2014 (47) (remove)
Document Type
- Doctoral Thesis (47) (remove)
Language
- English (47) (remove)
Has Fulltext
- yes (47)
Keywords
- Activity recognition (2)
- Wearable computing (2)
- Adaptive Data Structure (1)
- AhRR (1)
- Algorithm (1)
- Boosting (1)
- CYP1A1 (1)
- Classification (1)
- Closure (1)
- Code Generation (1)
Faculty / Organisational entity
- Kaiserslautern - Fachbereich Informatik (14)
- Kaiserslautern - Fachbereich Mathematik (13)
- Kaiserslautern - Fachbereich Maschinenbau und Verfahrenstechnik (5)
- Kaiserslautern - Fachbereich Sozialwissenschaften (5)
- Kaiserslautern - Fachbereich Chemie (4)
- Kaiserslautern - Fachbereich Elektrotechnik und Informationstechnik (3)
- Fraunhofer (ITWM) (1)
- Kaiserslautern - Fachbereich ARUBI (1)
- Kaiserslautern - Fachbereich Biologie (1)
Perceptual grouping is an integral part of visual object recognition. It organizes elements within our visual field according to a set of heuristics (grouping principles), most of which are not well understood. To identify their temporal processing dynamics (i.e., to identify whether they rely on neuronal feedforward or recurrent activation), we introduce the primed flanker task that is based on a firm empirical and theoretical background. In three sets of experiments, participants responded to visual stimuli that were either grouped by (1) similarity of brightness, shape, or size, (2) symmetry and closure, or (3) Good Gestalt. We investigated whether these grouping cues were effective in rapid visuomotor processing (i.e., in terms of response times, error rates, and priming effects) and whether the results met theory-driven indicators of feedforward processing. (1) In the first set of experiments with similarity cues, we varied subjective grouping strength and found that stronger grouping in the targets enhanced overall response times while stronger grouping in the primes enhanced priming effects in motor responses. We also obtained differences between rapid visuomotor processing and the subjective impression with cues of brightness and shape but not with cues of brightness and size. These results show that the primed flanker task is an objective measure for comparing different feedforward-transmitted groupings. (2) In the second set of experiments, we used the task to study grouping by symmetry and grouping by closure that are more complex than similarity cues. We obtained results that were mostly in accordance with a feedforward model. Some other factors (line of view, orientation of the symmetry axis) were irrelevant for processing of symmetry cues. Thus, these experiments suggest that closure and (possibly) viewpoint-independent symmetry cues are extracted rapidly during the first feedforward wave of neuronal processing. (3) In the third set of experiments, we used the task to study grouping by Good Gestalt (i.e., visual completion in occluded shapes). By varying the amount of occlusion, we found that the processing was in accordance with a feedforward model only when occlusion was very limited. Thus, these experiments suggest that Good Gestalt is not extracted rapidly during the first feedforward wave of neuronal processing but relies on recurrent activation. I conclude (1) that the primed flanker task is an excellent tool to identify and compare the processing characteristics of different grouping cues by behavioral means, (2) that grouping strength and other factors are strongly modulating these processing characteristics, which (3) challenges a dichotomous classification of grouping cues based on feedforward vs. recurrent processing (incremental grouping theory, Roelfsema, 2006), and (4) that a focus on temporal processing dynamics is necessary to understand perceptual grouping.
Mechanical ventilation of patients with severe lung injury is an important clinical treatment to ensure proper lung oxygenation and to mitigate the extent of collapsed lung regions. While current imaging technologies such as Computed Tomography (CT) and chest X-ray allow for a thorough inspection of the thorax, they are limited to static pictures and exhibit several disadvantages, including exposure to ionizing radiation and high cost. Electrical Impedance Tomography (EIT) is a novel method to determine functional processes inside the thorax such as lung ventilation and cardiac activity. EIT reconstructs the internal electrical conductivity distribution within the thorax from voltage measurements on the body surface. Conductivity changes correlate with important clinical parameters such as lung volume and perfusion. Current EIT systems and algorithms use simplified or generalized thorax models to solve the reconstruction problem, which reduce image quality and anatomical significance. In this thesis, the development of a clinically relevant workflow to compute sophisticated three-dimensional thorax models from patient-specific CT data is described. The method allows medical experts to generate a multi-material segmentation in an interactive and fast way, while a volumetric mesh is computed automatically from the segmentation. The significantly improved image quality and anatomical precision of EIT images reconstructed with these 3D models is reported, and the impact on clinical applicability is discussed. In addition, three projects concerning quantitative CT (qCT) measurements and multi-modal 3D visualization are presented, which demonstrate the importance and productivity of interdisciplinary research groups including computer scientists and medical experts. The results presented in this thesis contribute significantly to clinical research efforts to pave the way towards improved patient-specific treatments of lung injury using EIT and qCT.
Regular physical activity is essential to maintain or even improve an individual’s health. There exist various guidelines on how much individuals should do. Therefore, it is important to monitor performed physical activities during people’s daily routine in order to tell how far they meet professional recommendations. This thesis follows the goal to develop a mobile, personalized physical activity monitoring system applicable for everyday life scenarios. From the mentioned recommendations, this thesis concentrates on monitoring aerobic physical activity. Two main objectives are defined in this context. On the one hand, the goal is to estimate the intensity of performed activities: To distinguish activities of light, moderate or vigorous effort. On the other hand, to give a more detailed description of an individual’s daily routine, the goal is to recognize basic aerobic activities (such as walk, run or cycle) and basic postures (lie, sit and stand).
With recent progress in wearable sensing and computing the technological tools largely exist nowadays to create the envisioned physical activity monitoring system. Therefore, the focus of this thesis is on the development of new approaches for physical activity recognition and intensity estimation, which extend the applicability of such systems. In order to make physical activity monitoring feasible in everyday life scenarios, the thesis deals with questions such as 1) how to handle a wide range of e.g.
everyday, household or sport activities and 2) how to handle various potential users. Moreover, this thesis deals with the realistic scenario where either the currently performed activity or the current user is unknown during the development and training
phase of activity monitoring applications. To answer these questions, this thesis proposes and developes novel algorithms, models and evaluation techniques, and performs thorough experiments to prove their validity.
The contributions of this thesis are both of theoretical and of practical value. Addressing the challenge of creating robust activity monitoring systems for everyday life the concept of other activities is introduced, various models are proposed and validated. Another key challenge is that complex activity recognition tasks exceed the potential of existing classification algorithms. Therefore, this thesis introduces a confidence-based extension of the well known AdaBoost.M1 algorithm, called ConfAdaBoost.M1. Thorough experiments show its significant performance improvement compared to commonly used boosting methods. A further major theoretical contribution is the introduction and validation of a new general concept for the personalization of physical activity recognition applications, and the development of a novel algorithm (called Dependent Experts) based on this concept. A major contribution of practical value is the introduction of a new evaluation technique (called leave-one-activity-out) to simulate when performing previously unknown activities in a physical activity monitoring system. Furthermore, the creation and benchmarking of publicly available physical activity monitoring datasets within this thesis are directly benefiting the research community. Finally, the thesis deals with issues related to the implementation of the proposed methods, in order to realize the envisioned mobile system and integrate it into a full healthcare application for aerobic activity monitoring and support in daily life.
Optical character recognition (OCR) of machine printed text is ubiquitously considered as a solved problem. However, error free OCR of degraded (broken and merged) and noisy text is still challenging for modern OCR systems. OCR of degraded text with high accuracy is very important due to many applications in business, industry and large scale document digitization projects. This thesis presents a new OCR method for degraded
text recognition by introducing a combined ANN/HMM OCR approach. The approach
provides significantly better performance in comparison with state-of-the-art HMM based OCR methods and existing open source OCR systems. In addition, the thesis introduces novel applications of ANNs and HMMs for document image preprocessing and recognition of low resolution text. Furthermore, the thesis provides psychophysical experiments to determine the effect of letter permutation in visual word recognition of Latin and Cursive
script languages.
HMMs and ANNs are widely employed pattern recognition paradigms and have been
used in numerous pattern classification problems. This work presents a simple and novel method for combining the HMMs and ANNs in application to segmentation free OCR of degraded text. HMMs and ANNs are powerful pattern recognition strategies and their combination is interesting to improve current state-of-the-art research in OCR. Mostly, previous attempts in combining the HMMs and ANNs were focused on applying ANNs
as approximation of the probability density function or as a neural vector quantizer for HMMs. These methods either require combined NN/HMM training criteria [ECBG-MZM11] or they use complex neural network architecture like time delay or space displacement neural networks [BLNB95]. However, in this work neural networks are used as discriminative feature extractor, in combination with novel text line scanning mechanism, to extract discriminative features from unsegmented text lines. The features are
processed by HMMs to provide segmentation free text line recognition. The ANN/HMM modules are trained separately on a common dataset by using standard machine learning procedures. The proposed ANN/HMM OCR system also realizes to some extent several cognitive reading based strategies during the OCR. On a dataset of 1,060 degraded text lines extracted from the widely used UNLV-ISRI benchmark database [TNBC99], the presented system achieves a 30% reduction in error rate as compared to Google’s Tesseract OCR system [Smi13] and 43% reduction in error as compared to OCRopus OCR system [Bre08], which are the best open source OCR systems available today.
In addition, this thesis introduces new applications of HMMs and ANNs in OCR and document images preprocessing. First, an HMMs-based segmentation free OCR approach is presented for recognition of low resolution text. OCR of low resolution text is quite important due to presence of low resolution text in screen-shots, web images and video captions. OCR of low resolution text is challenging because of antialiased rendering and use of very small font size. The characters in low resolution text are usually joined to each other and they may appear differently at different locations on computer screen. This
work presents the use of HMMs in optical recognition of low resolution isolated characters and text lines. The evaluation of the proposed method shows that HMMs-based OCR techniques works quite well and reaches the performance of specialized approaches for OCR of low resolution text.
Then, this thesis presents novel applications of ANNs for automatic script recognition and orientation detection. Script recognition determines the written script on the page for the application of an appropriate character recognition algorithm. Orientation detection detects and corrects the deviation of the document’s orientation angle from the horizontal direction. Both, script recognition and orientation detection, are important preprocessing steps in developing robust OCR systems. In this work, instead of extracting handcrafted features, convolutional neural networks are used to extract relevant discriminative features for each classification task. The proposed method resulted in more than 95% script recognition accuracy on various multi-script documents at connected component level
and 100% page orientation detection accuracy for Urdu documents.
Human reading is a nearly analogous cognitive process to OCR that involves decoding of printed symbols into meanings. Studying the cognitive reading behavior may help in building a robust machine reading strategy. This thesis presents a behavioral study that deals on how cognitive system works in visual recognition of words and permuted non-words. The objective of this study is to determine the impact of overall word shape
in visual word recognition process. The permutation is considered as a source of shape degradation and visual appearance of actual words can be distorted by changing the constituent letter positions inside the words. The study proposes a hypothesis that reading of words and permuted non-words are two distinct mental level processes, and people use
different strategies in handling permuted non-words as compared to normal words. The hypothesis is tested by conducting psychophysical experiments in visual recognition of words from orthographically different languages i.e. Urdu, German and English. Experimental data is analyzed using analysis of variance (ANOVA) and distribution free rank tests to determine significance differences in response time latencies for two classes of data. The results support the presented hypothesis and the findings are consistent with
the dual route theories of reading.
Monte Carlo simulation is one of the commonly used methods for risk estimation on financial markets, especially for option portfolios, where any analytical approximation is usually too inaccurate. However, the usually high computational effort for complex portfolios with a large number of underlying assets motivates the application of variance reduction procedures. Variance reduction for estimating the probability of high portfolio losses has been extensively studied by Glasserman et al. A great variance reduction is achieved by applying an exponential twisting importance sampling algorithm together with stratification. The popular and much faster Delta-Gamma approximation replaces the portfolio loss function in order to guide the choice of the importance sampling density and it plays the role of the stratification variable. The main disadvantage of the proposed algorithm is that it is derived only in the case of Gaussian and some heavy-tailed changes in risk factors.
Hence, our main goal is to keep the main advantage of the Monte Carlo simulation, namely its ability to perform a simulation under alternative assumptions on the distribution of the changes in risk factors, also in the variance reduction algorithms. Step by step, we construct new variance reduction techniques for estimating the probability of high portfolio losses. They are based on the idea of the Cross-Entropy importance sampling procedure. More precisely, the importance sampling density is chosen as the closest one to the optimal importance sampling density (zero variance estimator) out of some parametric family of densities with respect to Kullback - Leibler cross-entropy. Our algorithms are based on the special choices of the parametric family and can now use any approximation of the portfolio loss function. A special stratification is developed, so that any approximation of the portfolio loss function under any assumption of the distribution of the risk factors can be used. The constructed algorithms can easily be applied for any distribution of risk factors, no matter if light- or heavy-tailed. The numerical study exhibits a greater variance reduction than of the algorithm from Glasserman et al. The use of a better approximation may improve the performance of our algorithms significantly, as it is shown in the numerical study.
The literature on the estimation of the popular market risk measures, namely VaR and CVaR, often refers to the algorithms for estimating the probability of high portfolio losses, describing the corresponding transition process only briefly. Hence, we give a consecutive discussion of this problem. Results necessary to construct confidence intervals for both measures under the mentioned variance reduction procedures are also given.
The heart is reported to show a net consumption of lactate. This may contribute up to 15% to the total body lactate disposal. In this work, the consumption of lactate was shown for the first
time on the single cell level with the new FRET-based lactate sensor Laconic.
Research published until today, almost exclusively reports the monocarboxylate transporter 1
(MCT1) as the transporter responsible for myocardial lactate uptake. As this membrane
transporter transports lactate together with H+ in a stoichiometry of 1:1, lactate transport is
coupled to pH regulation. Consequently, interactions of MCT1 and acid/base regulating proteins
(carbonic anhydrases (CAs and sodium bicarbonate co-transporters (NBCs)) are described in
the oocyte expression system, skeletal muscle and cancer cells.
In this work it is shown that activity of extracellular CA increases lactate uptake into mouse
cardiomyocytes by 27% and lactate induced JA/B by 42.8% to 46.2%. This effect is most likely
mediated via NBC/CA interaction because inhibition of extracellular CA reduces HCO3--
dependent acid extruding JA/B by 53.3% to 78.4%. This may link lactate uptake to cellular
respiration. When lactate was applied in medium gassed with 100% N2, lactate induced
acidification was 12.6% faster than in medium gassed with 100% O2. Thus, CO2 produced on
the pathway transferring redox energy from substrates like glucose and lactate to ADP and
phosphate via oxidative phosphorylation, may support further lactate uptake. The findings of
this work suggest an auto regulation of lactate uptake via CO2 release in ventricular mouse
cardiomyocytes.
In this thesis we studied and investigated a very common but a long existing noise problem and we provided a solution to this problem. The task is to deal with different types of noise that occur simultaneously and which we call hybrid. Although there are individual solutions for specific types one cannot simply combine them because each solution affects the whole speech. We developed an automatic speech recognition system DANSR ( Dynamic Automatic Noisy Speech Recognition System) for hybrid noisy environmental noise. For this we had to study all of speech starting from the production of sounds until their recognition. Central elements are the feature vectors on which pay much attention. As an additional effect we worked on the production of quantities for psychoacoustic speech elements.
The thesis has four parts:
1) The first part we give an introduction. The chapter 2 and 3 give an overview over speech generation and recognition when machines are used. Also noise is considered.
2) In the second part we describe our general system for speech recognition in a noisy environment. This is contained in the chapters 4-10. In chapter 4 we deal with data preparation. Chapter 5 is concerned with very strong noise and its modeling using Poisson distribution. In the chapters 5-8 we deal with parameter based modeling. Chapter 7 is concerned with autoregressive methods in relation to the vocal tract. In the chapters 8 and 9 we discuss linear prediction and its parameters. Chapter 9 is also concerned with quadratic errors, the decomposition into sub-bands and the use of Kalman filters for non-stationary colored noise in chapter 10. There one finds classical approaches as long we have used and modified them. This includes covariance mehods, the method of Burg and others.
3) The third part deals firstly with psychoacoustic questions. We look at quantitative magnitudes that describe them. This has serious consequences for the perception models. For hearing we use different scales and filters. In the center of the chapters 12 and 13 one finds the features and their extraction. The fearures are the only elements that contain information for further use. We consider here Cepstrum features and Mel frequency cepstral coefficients(MFCC), shift invariant local trigonometric transformed (SILTT), linear predictive coefficients (LPC), linear predictive cepstral coefficients (LPCC), perceptual linear predictive (PLP) cepstral coefficients. In chapter 13 we present our extraction methods in DANSR and how they use window techniques And discrete cosine transform (DCT-IV) as well as their inverses.
4) The fourth part considers classification and the ultimate speech recognition. Here we use the hidden Markov model (HMM) for describing the speech process and the Gaussian mixture model (GMM) for the acoustic modelling. For the recognition we use forward algorithm, the Viterbi search and the Baum-Welch algorithm. We also draw the connection to dynamic time warping (DTW). In the rest we show experimental results and conclusions.
In this thesis, we combine Groebner basis with SAT Solver in different manners.
Both SAT solvers and Groebner basis techniques have their own strength and weakness.
Combining them could fix their weakness.
The first combination is using Groebner techniques to learn additional binary clauses for SAT solver from a selection of clauses. This combination is first proposed by Zengler and Kuechlin.
However, in our experiments, about 80 percent Groebner basis computations give no new binary clauses.
By selecting smaller and more compact input for Groebner basis computations, we can significantly
reduce the number of inefficient Groebner basis computations, learn much more binary clauses. In addition,
the new strategy can reduce the solving time of a SAT Solver in general, especially for large and hard problems.
The second combination is using all-solution SAT solver and interpolation to compute Boolean Groebner bases of Boolean elimination ideals of a given ideal. Computing Boolean Groebner basis of the given ideal is an inefficient method in case we want to eliminate most of the variables from a big system of Boolean polynomials.
Therefore, we propose a more efficient approach to handle such cases.
In this approach, the given ideal is translated to the CNF formula. Then an all-solution SAT Solver is used to find the projection of all solutions of the given ideal. Finally, an algorithm, e.g. Buchberger-Moeller Algorithm, is used to associate the reduced Groebner basis to the projection.
We also optimize the Buchberger-Moeller Algorithm for lexicographical ordering and compare it with Brickenstein's interpolation algorithm.
Finally, we combine Groebner basis and abstraction techniques to the verification of some digital designs that contain complicated data paths.
For a given design, we construct an abstract model.
Then, we reformulate it as a system of polynomials in the ring \({\mathbb Z}_{2^k}[x_1,\dots,x_n]\).
The variables are ordered in a way such that the system has already been a Groebner basis w.r.t lexicographical monomial ordering.
Finally, the normal form is employed to prove the desired properties.
To evaluate our approach, we verify the global property of a multiplier and a FIR filter using the computer algebra system Singular. The result shows that our approach is much faster than the commercial verification tool from Onespin on these benchmarks.
‘Dioxin-like’ (DL) compounds occur ubiquitously in the environment. Toxic responses associated with specific dibenzo-p-dioxins (PCDDs), dibenzofurans (PCDFs), and polychlorinated biphenyls (PCBs) include dermal toxicity, immunotoxicity, liver toxicity, carcinogenicity, as well as adverse effects on reproduction, development, and endocrine functions. Most, if not all of these effects are believed to be due to interaction of these compounds with the aryl hydrocarbon receptor (AhR).
With tetrachlorodibenzo-p-dioxin (TCDD) as representatively most potent congener, a toxic equivalency factor (TEF) concept was employed, in which respective congeners were assigned to a certain TEF-value reflecting the compound’s toxicity relative to TCDD’s.
The EU-project ‘SYSTEQ’ aimed to develop, validate, and implement human systemic TEFs as indicators of toxicity for DL-congeners. Hence, the identification of novel quantifiable biomarkers of exposure was a major objective of the SYSTEQ project.
In order to approach to this objective, a mouse whole genome microarray analysis was applied using a set of seven individual congeners, termed the ‘core congeners’. These core congeners (TCDD, 1-PeCDD, 4-PeCDF, PCB 126, PCB 118, PCB 156, and the non dioxin-like PCB 153), which contribute to approximately 90% of toxic equivalents (TEQs) in the human food chain, were further tested in vivo as well as in vitro. The mouse whole genome microarray revealed a conserved list of differentially regulated genes and pathways associated with ‘dioxin-like’ effects.
A definite data-set of in vitro studies was supposed to function as a fundament for a probable establishment of novel TEFs. Thus, CYP1A induction measured by EROD activity, which represents a sensitive and yet best known marker for dioxin-like effects, was used to estimate potency and efficacy of selected congeners. For this study, primary rat hepatocytes and the rat hepatoma cell line H4IIE were used as well as the core congeners and an additional group of compounds of comparable relevance for the environment: 1,6-HxCDD, 1,4,6-HpCDD, TCDF, 1,4-HxCDF, 1,4,6-HpCDF, PCB 77, and PCB 105.
Besides, a human whole genome microarray experiment was applied in order to gain knowledge with respect to TCDD’s impact towards cells of the immune system. Hence, human primary blood mononuclear cells (PBMCs) were isolated from individuals and exposed to TCDD or to TCDD in combination with a stimulus (lipopolysaccharide (LPS), or phytohemagglutinin (PHA)). A few members of the AhR-gene batterie were found to be regulated, and minor data with respect to potential TCDD-mediated immunomodulatory effects were given. Still, obtained data in this regard was limited due to great inter-individual differences.
This thesis is devoted to the modeling and simulation of Asymmetric Flow Field Flow Fractionation, which is a technique for separating particles of submicron scale. This process is a part of large family of Field Flow Fractionation techniques and has a very broad range of industrial applications, e. g. in microbiology, chemistry, pharmaceutics, environmental analysis.
Mathematical modeling is crucial for this process, as due to the own nature of the process, lab ex- periments are difficult and expensive to perform. On the other hand, there are several challenges for the mathematical modeling: huge dominance (up to 106 times) of the flow over the diffusion, highly stretched geometry of the device. This work is devoted to developing fast and efficient algorithms, which take into the account the challenges, posed by the application, and provide reliable approximations for the quantities of interest.
We present a new Multilevel Monte Carlo method for estimating the distribution functions on a compact interval, which are of the main interest for Asymmetric Flow Field Flow Fractionation. Error estimates for this method in terms of computational cost are also derived.
We optimize the flow control at the Focusing stage under the given constraints on the flow and present an important ingredients for the further optimization, such as two-grid Reduced Basis method, specially adapted for the Finite Volume discretization approach.