The recognition of day-to-day activities is still a very challenging and important research topic. During recent years, a lot of research has gone into designing and realizing smart environ- ments in different application areas such as health care, maintenance, sports or smart homes. As a result, a large amount of sensor modalities were developed, different types of activity and context recognition services were implemented and the resulting systems were benchmarked using state-of-the-art evaluation techniques. However, so far hardly any of these approaches have found their way into the market and consequently into the homes of real end-users on a large scale. The reason for this is, that almost all systems have one or more of the following characteristics in common: expensive high-end or prototype sensors are used which are not af- fordable or reliable enough for mainstream applications; many systems are deployed in highly instrumented environments or so-called "living labs", which are far from real-life scenarios and are often evaluated only in research labs; almost all systems are based on complex system con- figurations and/or extensive training data sets, which means that a large amount of data must be collected in order to install the system. Furthermore, many systems rely on a user and/or environment dependent training, which makes it even more difficult to install them on a large scale. Besides, a standardized integration procedure for the deployment of services in existing environments and smart homes has still not been defined. As a matter of fact, service providers use their own closed systems, which are not compatible with other systems, services or sensors. It is clear, that these points make it nearly impossible to deploy activity recognition systems in a real daily-life environment, to make them affordable for real users and to deploy them in hundreds or thousands of different homes.
This thesis works towards the solution of the above mentioned problems. Activity and context recognition systems designed for large-scale deployment and real-life scenarios are intro- duced. Systems are based on low-cost, reliable sensors and can be set up, configured and trained with little effort, even by technical laymen. It is because of these characteristics that we call our approach "minimally invasive". As a consequence, large amounts of training data, that are usu- ally required by many state-of-the-art approaches, are not necessary. Furthermore, all systems were integrated unobtrusively in real-world/similar to real-world environments and were evalu- ated under real-life, as well as similar to real-life conditions. The thesis addresses the following topics: First, a sub-room level indoor positioning system is introduced. The system is based on low-cost ceiling cameras and a simple computer vision tracking approach. The problem of user identification is solved by correlating modes of locomotion patterns derived from the trajectory of unidentified objects and on-body motion sensors. Afterwards, the issue of recognizing how and what mainstream household devices have been used for is considered. Based on a low-cost microphone, the water consumption of water-taps can be approximated by analyzing plumbing noise. Besides that, operating modes of mainstream electronic devices were recognized by using rule-based classifiers, electric current features and power measurement sensors. As a next step, the difficulty of spotting subtle, barely distinguishable hand activities and the resulting object interactions, within a data set containing a large amount of background data, is addressed. The problem is solved by introducing an on-body core system which is configured by simple, one-time physical measurements and minimal data collections. The lack of large training sets is compensated by fusing the system with activity and context recognition systems, that are able to reduce the search space observed. Amongst other systems, previously introduced approaches and ideas are revisited in this section. An in-depth evaluation shows the impact of each fusion procedure on the performance and run-time of the system. The approaches introduced are able to provide significantly better results than a state-of-the-art inertial system using large amounts of training data. The idea of using unobtrusive sensors has also been applied to the field of behavior analysis. Integrated smartphone sensors are used to detect behavioral changes of in- dividuals due to medium-term stress periods. Behavioral parameters related to location traces, social interactions and phone usage were analyzed to detect significant behavioral changes of individuals during stressless and stressful time periods. Finally, as a closing part of the the- sis, a standardization approach related to the integration of ambient intelligence systems (as introduced in this thesis) in real-life and large-scale scenarios is shown.
The hypervolume subset selection problem consists of finding a subset, with a given cardinality, of a nondominated set of points that maximizes the hypervolume indicator. This problem arises in selection procedures of population-based heuristics for multiobjective optimization, and for which practically efficient algorithms are strongly required. In this article, we provide two new formulations for the two-dimensional variant of this problem.
The first is an integer programming formulation that can be solved by solving its linear relaxation. The second formulation is a \(k\)-link shortest path formulation on a special digraph with Monge property that can be solved by dynamic programming in \(\mathcal{O}(n^2)\) time complexity. This improves upon the existing result of \(O(n^3)\) in Bader.
Edgeworth expansions have been introduced as a generalization of the central limit theorem and allow to investigate the convergence properties of sums of i.i.d. random variables. We consider triangular arrays of lattice random vectors and obtain a valid Edgeworth expansion for this case. The presented results can be used, for example, to study the convergence behavior of lattice models.
We propose a model for acid-mediated tumor invasion involving two different scales: the microscopic one, for the dynamics of intracellular protons and their exchange with their extracellular counterparts, and the macroscopic scale of interactions between tumor cell and normal cell populations, along with the evolution of extracellular protons. We also account for the tactic behavior of cancer cells, the latter being assumed to biase their motion according to a gradient of extracellular protons (following [2,31] we call this pH taxis). A time dependent (and also time delayed) carrying capacity for the tumor cells in response to the effects of acidity is considered as well. The global well posedness of the resulting multiscale model is proved with a regularization and fixed point argument. Numerical simulations are performed in order to illustrate the behavior of the model.
This thesis is devoted to the computational aspects of intersection theory and enumerative geometry. The first results are a Sage package Schubert3 and a Singular library schubert.lib which both provide the key functionality necessary for computations in intersection theory and enumerative geometry. In particular, we describe an alternative method for computations in Schubert calculus via equivariant intersection theory. More concretely, we propose an explicit formula for computing the degree of Fano schemes of linear subspaces on hypersurfaces. As a special case, we also obtain an explicit formula for computing the number of linear subspaces on a general hypersurface when this number is finite. This leads to a much better performance than classical Schubert calculus.
Another result of this thesis is related to the computation of Gromov-Witten invariants. The most powerful method for computing Gromov-Witten invariants is the localization of moduli spaces of stable maps. This method was introduced by Kontsevich in 1995. It allows us to compute Gromov-Witten invariants via Bott's formula. As an insightful application, we computed the numbers of rational curves on general complete intersection Calabi-Yau threefolds in projective spaces up to degree six. The results are all in agreement with predictions made from mirror symmetry.
According to the domain specific models of speech perception, speech is supposed to be processed distinctively compared to non-speech. This assumption is supported by many studies dealing with the processing of speech and non-speech stimuli. However, the complexity of both stimulus classes is not matched in most studies, which might be a confounding factor, according to the cue specific models of speech perception. One solution is spectrally rotated speech, which has already been used in a range of fMRI and PET studies. In order to be able to investigate the role of stimulus complexity, vowels, spectrally rotated vowels and a second non-speech condition with two bands of sinusoidal waves, representing the first two formants of the vowels, were used in the present thesis. A detailed description of the creation and the properties of the whole stimulus set are given in Chapter 2 (Experiment 1) of this work. These stimuli were used to investigate the auditory processing of speech and non-speech sounds in a group of dyslexic adults and age matched controls (Experiment 2). The results support the assumption of a general auditory deficit in dyslexia. In order to compare the sensory processing of speech and non-speech in healthy adults on the electrophysiological level, stimuli were also presented within a multifeature oddball paradigm (Experiment 3). Vowels evoked a larger mismatch negativity (MMN) compared to both non-speech stimulus types. The MMN evoked by tones and spectrally rotated tones were compared in Experiment 4, to investigate the role of harmony. No difference in the area of MMN was found, indicating that the results found in Experiment 3 were not moderated by the harmonic structure of the vowels. All results are discussed in the context of the domain and cue specific models of speech perception.
An huge amount of computational models and programming languages have been proposed
for the description of embedded systems. In contrast to traditional sequential programming
languages, they cope directly with the requirements for embedded systems: direct support for
concurrent computations and periodic interaction with the environment are only some of the
features they offer. Synchronous languages are one class of languages for the development of
embedded systems and they follow the fundamental principle that the execution is divided into
a sequence of logical steps. Thereby, each step follows the simplification that the computation
of the outputs is finished directly when the inputs are available. This rigorous abstraction leads
to well-defined deterministic parallel composition in general, and to deterministic abortion
and suspension in imperative synchronous languages in particular. These key features also
allow to translate programs to hardware and software, and also formal verification techniques
like model checking can be easily applied.
Besides the advantages of imperative synchronous languages, also some drawbacks can
be listed. Over-synchronization is an effect being caused by parallel threads which have to
synchronize for each execution step, even if they do not communicate, since the synchronization
is implicitly forced by the control-flow. This thesis considers the idea of clock refinement to
introduce several abstraction layers for communication and synchronization in addition to the
existing single-clock abstraction. Thereby, clocks can be refined by several independent clocks
so that a controlled amount of asynchrony between subsequent synchronization points can be
exploited by compilers. The declarations of clocks form a tree, and clocks can be defined within
the threads of the parallel statement, which allows one to do independent computations based
on these clocks without synchronizing the threads. However, the synchronous abstraction is
kept at each level of the abstraction.
Clock refinement is introduced in this thesis as an extension to the imperative synchronous
language Quartz. Therefore, new program statements are introduced which allow to define
a new clock as a refinement of an existing one and to finish a step based on a certain clock.
Examples are considered to show the impact of the behavior of the new statements to
the already existing statements, before the semantics of this extension is formally defined.
Furthermore, the thesis presents a compile algorithm to translate programs to an intermediate
format, and to translate the intermediate format to a hardware description. The advantages
obtained by the new modeling feature are finally evaluated based on examples.
From an evolutionary point of view, it can be assumed that visual processing and rapid detection of potentially dangerous stimuli in the environment (e.g., perilous animals) is highly adaptive for all humans. In the present dissertation, I address three research questions; (1) Is information processing of threatening stimuli enhanced in individuals with specific phobias? (2) Are there any differences between the different types of phobia (e.g., spider phobia vs. snake phobia)? (3) Is the frequently reported attentional bias of individuals with specific phobias - which may contribute to an enhancement in information processing – also detectable in a prior entry paradigm? In Experiments 1 to 3 of the present thesis non-anxious control, spider-fearful, snake-fearful, and blood-injection-injury-fearful participants took part in the study. We applied in each experiment a response priming paradigm which has a strong theoretical (cf. rapid-chase theory; Schmidt, Niehaus, & Nagel, 2006; Schmidt, Haberkamp, Veltkamp et al., 2011) as well as empirical background (cf. Schmidt, 2002). We show that information processing in fearful individuals is indeed enhanced for phobic images (i.e., spiders for spider-fearful participants; injuries for blood-injury-injection(BII)-fearful individuals). However, we found marked differences between the different types of phobia. In Experiment 1 and 2 (Chapter 2 and 3), spiders had a strong and specific influence in the group of spider-fearful individuals: Phobic primes entailed the largest priming effects, and phobic targets accelerated responses, both effects indicating speeded response activation by phobic images. In snake-fearful participants (Experiment 1, Chapter 2), this processing enhancement for phobic material was less pronounced and extended to both snake and spider images. In Experiment 3 (Chapter 4), we demonstrated that early information processing for pictures of small injuries is also enhanced in BII-fearful participants, even though BII fear is unique in that BII-fearful individuals show opposite physiological reactions when confronted with the phobic stimulus compared to individuals with animal phobias. These results show that already fast visuomotor responses are further enhanced in spider- and BII-fearful participants. Results give evidence that responses are based on the first feedforward sweep of neuronal activation proceeding through the visuomotor system. I propose that the additional enhancement in spider- and BII-fearful individuals depend on a specific hardwired binding of elementary features belonging to the phobic object in fearful individuals (i.e., effortless recognition of the respective phobic object via hardwired neuronal conjunctions). I suggest that these hardwired conjunctions developed due to long-term perceptual learning processes. We also investigate the frequently reported attentional bias of phobic individuals and showed that this bias is detectable in temporal order judgments using a prior entry paradigm. I assume that perceptual learning processes might also strengthen the attentional bias, for example, by providing a more salient bottom-up signal that draws attention involuntarily. In sum, I conclude that (1) early information processing of threatening stimuli is indeed enhanced in individuals with specific phobias but that (2) differences between divers types of phobia exist (i.e., spider- and BII-fearful participants show enhanced information of the respective phobic object; though, snake-fearful participants show no specific information processing enhancement of snakes); (3) the frequently reported attentional bias of spider-fearful individuals is also detectable in a prior entry paradigm.
In 2006 Jeffrey Achter proved that the distribution of divisor class groups of degree 0 of function fields with a fixed genus and the distribution of eigenspaces in symplectic similitude groups are closely related to each other. Gunter Malle proposed that there should be a similar correspondence between the distribution of class groups of number fields and the distribution of eigenspaces in ceratin matrix groups. Motivated by these results and suggestions we study the distribution of eigenspaces corresponding to the eigenvalue one in some special subgroups of the general linear group over factor rings of rings of integers of number fields and derive some conjectural statements about the distribution of \(p\)-parts of class groups of number fields over a base field \(K_{0}\). Where our main interest lies in the case that \(K_{0}\) contains the \(p\)th roots of unity, because in this situation the \(p\)-parts of class groups seem to behave in an other way like predicted by the popular conjectures of Henri Cohen and Jacques Martinet. In 2010 based on computational data Malle has succeeded in formulating a conjecture in the spirit of Cohen and Martinet for this case. Here using our investigations about the distribution in matrixgroups we generalize the conjecture of Malle to a more abstract level and establish a theoretical backup for these statements.
In this paper we analyze the vibrations of nonlinear structures by means of the novel approach of isogeometric finite elements. The fundamental idea of isogeometric finite elements is to apply the same functions, namely B-Splines and NURBS (Non-Uniform Rational B-Splines), for describing the geometry and for representing the numerical solution. In case of linear vibrational analysis, this approach has already been shown to possess substantial advantages over classical finite elements, and we extend it here to a nonlinear framework based on the harmonic balance principle.
As application, the straight nonlinear Euler-Bernoulli beam is used, and overall, it is demonstrated that isogeometric finite elements with B-Splines in combination with the harmonic balance method are a powerful means for the analysis of nonlinear structural vibrations. In particular, the smoother k-method provides higher accuracy than the p-method for isogeometric nonlinear vibration analysis.
Multi-Modal Activity Recognition Systems with Minimal Training Data and Unobtrusive Environmental Instrumentations
(2014)
