Search

Human action recognition in challenging video data is becoming an increasingly important research area. Given the growing number of cameras and robots pointing their lenses at humans, the need for automatic recognition of human actions arises, promising Google-style video search and automatic video summarisation/description. Furthermore, for any autonomous robotic system to interact with humans, it must rst be able to understand and quickly react to human actions. Although the best action classication methods aggregate features from the entire video clip in which the action unfolds, this global representation may include irrelevant scene context and movements which are shared amongst multiple action classes. For example, a waving action may be performed whilst walking, however if the walking movement appears in distinct action classes, then it should not be included in training a waving movement classier. For this reason, we propose an action classication framework in which more discriminative action subvolumes are learned in a weakly supervised setting, owing to the diculty of manually labelling massive video datasets. The learned models are used to simultaneously classify video clips and to localise actions to a given space-time subvolume. Each subvolume is cast as a bag-of-features (BoF) instance in a multiple-instance-learning framework, which in turn is used to learn its class membership. We demonstrate quantitatively that even with single xed-sized subvolumes, the classication performance of our proposed algorithm is superior to our BoF baseline on the majority of performance measures, and shows promise for space-time action localisation on the most challenging video datasets. Exploiting spatio-temporal structure in the video should also improve results, just as deformable part models have proven highly successful in object recognition. However, whereas objects have clear boundaries which means we can easily dene a ground truth for initialisation, 3D space-time actions are inherently ambiguous and expensive to annotate in large datasets. Thus, it is desirable to adapt pictorial star models to action datasets without location annotation, and to features invariant to changes in pose such as bag-of-feature and Fisher vectors, rather than low-level HoG. Thus, we propose local deformable spatial bag-of-features (LDSBoF) in which local discriminative regions are split into axed grid of parts that are allowed to deform in both space and time at test-time. In our experimental evaluation we demonstrate that by using local, deformable space-time action parts, we are able to achieve very competitive classification performance, whilst being able to localise actions even in the most challenging video datasets. A recent trend in action recognition is towards larger and more challenging datasets, an increasing number of action classes and larger visual vocabularies. For the global classication of human action video clips, the bag-of-visual-words pipeline is currently the best performing. However, the strategies chosen to sample features and construct a visual vocabulary are critical to performance, in fact often dominating performance. Thus, we provide a critical evaluation of various approaches to building a vocabulary and show that good practises do have a signicant impact. By subsampling and partitioning features strategically, we are able to achieve state-of-the-art results on 5 major action recognition datasets using relatively small visual vocabularies. Another promising approach to recognise human actions first encodes the action sequence via a generative dynamical model. However, using classical distances for their classication does not necessarily deliver good results. Therefore we propose a general framework for learning distance functions between dynamical models, given a training set of labelled videos. The optimal distance function is selected among a family of `pullback' ones, induced by a parametrised mapping of the space of models. We focus here on hidden Markov models and their model space, and show how pullback distance learning greatly improves action recognition performances with respect to base distances. Finally, the action classication systems that use a single global representation for each video clip are tailored for oine batch classication benchmarks. For human-robot interaction however, current systems fall short, either because they can only detect one human action per video frame, or because they assume the video is available ahead of time. In this work we propose an online human action detection system that can incrementally detect multiple concurrent space-time actions. In this way, it becomes possible to learn new action classes on-the-fly, allowing multiple people to actively teach and interact with a robot.

Teaching reading in the mainstream classroom is a challenge. This challenge is compounded when trying to meet the needs of English language learners. Recently, Response to Intervention (RTI) has been suggested as a framework for classroom teachers to use in order to meet the wide range of needs of their students. The purpose of this study was to evaluate the effectiveness of an RTI reading instructional model when used with English language learners (ELLs) in mainstream classrooms. A mixed-method, quasi-experimental, pre/post design was implemented. A purposive sample was drawn from third grade students in schools that integrated ELLs and native English speakers. Dynamic Indicators of Basic Early Literacy Skills (DIBELS) and the Developmental Reading Assessment (DRA) were used to assess Reading achievement. Student reading assessment scores from the end of second grade were used as a covariate. Data were analyzed using MANCOVA. Further analysis of the ELLs included in the study was conducted using MANCOVA with Assessing Comprehension and Communication in English State-to-State (ACCESS) reading subsections as a covariate. The results of this study showed that ELLs in schools implementing reading RTI were able to achieve predicted reading benchmark levels at the same rate as native English speakers in the same schools. Overall, the number of students reaching benchmark levels was nearly equal in RTI and comparison schools; however, RTI schools had a greater number of ELLs reaching the benchmark. ELLs in RTI schools made the greatest gains in reading scores between assessments as compared to all other groups in the study. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Chemical warfare agents are a toxic class of compounds that are incredibly harmful to human health. Methods of detoxification and decontamination currently exist, however they all suffer from problems that involve logistical transport or involve technologies that directly address liquid threats instead of vapors. One promising method of detoxification involves the oxidation of these compounds into less-harmful species. The relatively large chemical size and complexity of modern-day chemical warfare agents, however, precludes a straightforward analysis of the chemical transformations that take place on novel decontaminating materials. Additionally, a fundamental understanding of reaction mechanisms that occur on novel material surfaces is required before improved materials can be developed. To this end, the oxidation of three simpler, smaller organic molecules were studied over a variety of materials in order to build up a chemical understanding of the systems under study. The photoepoxidation of propene into propene oxide was observed to readily occur over an in-house developed dual titania-silica catalyst created by atomic layer deposition. The subsequent photoinduced degradation of produced propene oxide was observed to occur over the novel catalyst. Next, the oxidation of CO was studied over a Pt/TiO2 catalyst while in the presence of humidity. The addition of water was shown to enable an alternative, low energy pathway that closely followed the water gas shift, but ended upon the production of stable surface-bound formates. Gaseous oxygen was found to subsequently oxidize these surface formates into the full oxidation product, CO2. Next, the oxidation of methanol was studied over the same Pt/TiO2 catalyst. It was discovered that the water produced when methanol initially adsorbs to the catalyst surface is responsible for unlocking the oxidative capacity of the material. Finally, a custom packedbed reactor was designed and built that enabled unique experimental capabilities not yet available in commercial systems, and will be used in the future to directly test the oxidative capabilities of novel materials for chemical warfare agent destruction.

The polarisation properties of light are often used as a means of information transfer, as well as in assessing the structural and compositional properties of materials. In many practical situations, however, the propagation of polarised light is impeded by random scattering, which tends to scramble the information content of the underlying fields. Theoretical study of the random scattering of light is an old, complex field of research, in which approximate, numerical methods are often favoured over exact mathematical analysis. Random matrix theory, in which systems are modelled using random scattering matrices, has uncovered universal properties of wide classes of random media, most notably in the field of quantum scattering. Despite also finding success in optics, a random matrix theory of polarised light has yet to be pursued. In this thesis we apply the notion of random matrices to develop statistical techniques for modelling the random scattering of polarised light. We present a full derivation of the symmetries of the vectorial scattering and transfer matrices that describe dielectric scattering media, including the scattering of evanescent wave components. We then consider the circular ensembles as a simple random matrix model and explore its implications for the scattering of polarised light. Moving beyond elementary models, a rigorous, statistical theory of the scattering matrix for discrete random media is presented, and exact mathematical results are derived in certain special cases. A numerical simulation method for studying scattering matrices describing random media of arbitrary thickness is then developed and validated against known physical phenomena. Finally, the techniques developed within this thesis are applied to the problem of the recovery of polarisation information within light that has propagated through a random medium.

The development of executive function (EF) has garnered attention in recent years because of its association with many positive outcomes. Although evidence suggests these skills can be trained, little research has focused on the processes that promote strong EF skills in high-risk children. Study 1 sought to investigate spontaneous self-regulatory strategies in 138 4- to 7-year-old homeless children during an EF task and to understand the relation between verbal and physical strategies, performance on EF tasks, and school outcomes. As hypothesized, results indicated that physical strategy use was significantly related to general EF, and that EF mediated the relation between physical strategy use and academic achievement and peer competence at school. Study 2 sought to investigate whether similar strategies can be trained and are related to performance on a delay task. 106 4- to 7- year-old homeless children were randomly assigned to training and control groups, and performance on two delay tasks was examined. As expected, children in the training group displayed significantly more strategies on the training task than did children in the control group and that these strategies were significantly related to task performance. However, there were no overall group differences in performance. Exploratory analyses revealed some evidence for a significant relation between training and performance only for older children. Additionally, results demonstrated some transfer of trained strategies to a generalization task, although these were not related to performance. Overall, evidence indicates some potential benefit of training children to use strategies during delay tasks, with implications for interventions aimed at promoting EF development and long-term school success.

A novel fabrication technique to print colloidal single-nanoparticle assemblies onto flat surfaces was designed. The mean printing accuracy was found to be 60.8nm and printing yield was found to be 98.1%. The novel technique was utilised to print nanoparticles on an array of surfaces. Changes in the spectral position of peaks in the dark field scattering spectrum of these nanoparticles were indicative of strong optical coupling between nanoparticles and surfaces, likely owing to the lack of adhesion layer between nanoparticle and surface. These changes were supported and further illustrated through computer simulations. Nanocubes printed via this technique were utilised for in situ SERS measurements of hot electron catalysis of halogenated thiophenols. Hot electron generation was also utilised to selectively cleave thiol-bonds on the surface of nanoantennas for nanoparticle placement on gold drop-tie antennas.

Modern single photon quantum technologies require on-demand and deterministic sources. Single photon emission has been observed from various solid state systems, such as individual molecules, defects and colour centres in diamond and 2D materials, and artificial atoms - quantum dots. However, the quantum emitters suffer from the slow radiation rate and omnidirectional emission, preventing their practical applications. First, this Thesis proposes to exploit the charged excitonic states in individual colloidal quantum dots for the enhancement of radiation rate. Electron injection to a quantum dot increases the number of de-excitation pathways, hence boosting the single photon generation speed. The optical properties of charged excitons depend drastically on the number of injected electrons, which can be controlled by applying a voltage bias in an electrochemical cell, thus providing for an active and deterministic way to manipulate the single photon emission. The charge transfer allows for a deterministic manipulation of quantum dot photodynamics, with an observed 210-fold increase of the radiation rate, accompanied by a 12-fold decrease of the emission intensity, all while preserving single-photon emission characteristics. Secondly, this Thesis proposes a 3D metal-dielectric parabolic antenna with an individual quantum dot in its focal point as a source of collimated single photons which can then be easily extracted and manipulated. Compared to conventional nano-antennas, 3D parabolic antenna design does not require near-field coupling, hence it is very robust against misalignment issues, and minimally affected by absorption in the metal. The parabolic antenna provides one of the largest reported experimental directivity (D=106) and the lowest beam divergence (13.5 deg), a broadband operation over all the visible and near-IR range, together with more than 96% extraction efficiency, offering a practical advantage for quantum technological applications.

Teaching reading in the mainstream classroom is a challenge. This challenge is compounded when trying to meet the needs of English language learners. Recently, Response to Intervention (RTI) has been suggested as a framework for classroom teachers to use in order to meet the wide range of needs of their students. The purpose of this study was to evaluate the effectiveness of an RTI reading instructional model when used with English language learners (ELLs) in mainstream classrooms. A mixed-method, quasi-experimental, pre/post design was implemented. A purposive sample was drawn from third grade students in schools that integrated ELLs and native English speakers. Dynamic Indicators of Basic Early Literacy Skills (DIBELS) and the Developmental Reading Assessment (DRA) were used to assess Reading achievement. Student reading assessment scores from the end of second grade were used as a covariate. Data were analyzed using MANCOVA. Further analysis of the ELLs included in the study was conducted using MANCOVA with Assessing Comprehension and Communication in English State-to-State (ACCESS) reading subsections as a covariate. The results of this study showed that ELLs in schools implementing reading RTI were able to achieve predicted reading benchmark levels at the same rate as native English speakers in the same schools. Overall, the number of students reaching benchmark levels was nearly equal in RTI and comparison schools; however, RTI schools had a greater number of ELLs reaching the benchmark. ELLs in RTI schools made the greatest gains in reading scores between assessments as compared to all other groups in the study.

Promiscuous mutant EcoRI endonucleases produce lethal to sub-lethal effects because they cleave E. coli DNA despite the presence of the EcoRI methylase. Three promiscuous mutant forms, Ala138Thr, Glu192Lys and His114Tyr, have been characterized with respect to their binding affinities and first-order cleavage rate constants towards the three classes of DNA sites: specific, miscognate (EcoRI*) and nonspecific. We have made the unanticipated and counterintuitive observations that the mutant endonucleases that exhibit relaxed specificity in vivo nevertheless bind more tightly than the wild-type enzyme to the specific recognition sequence in vitro and show even greater preference for binding to the cognate GAATTC site over miscognate sites. Binding preference for EcoRI* over nonspecific DNA is also improved. The mutant enzymes cleave the cognate site GAATTC at a normal rate, but cleave EcoRI* sites faster than does the wild-type enzyme. Thus, the mutant enzymes use two mechanisms to partially bypass the multiple fail-safe mechanisms that protect against cleavage of genomic DNA in cells carrying the wild-type EcoRI restriction-modification system: (a) Binding to EcoRI* sites is more probable than for wild-type enzyme because nonspecific DNA is less effective as a competitive inhibitor; (b) The combination of increased affinity and faster cleavage at EcoRI* sites makes double-strand cleavage of these sites a more probable outcome than it is for the wild-type enzyme. The crystal structure of the A138T "promiscuous" mutant enzyme in complex with specific DNA shows reveals no changes in protein contacts to the bases of the GAATTC site relative to the wild-type complex; however, there are changes in water-mediated contacts between the enzyme and flanking bases, and changes in protein-phosphate contacts. These observations lead us to hypothesize that the improved specific DNA binding of the A138T enzyme relative to the wild-type enzyme is not attributable to a single new protein DNA contact, but rather the distributed effect of the improved complementarity between the mutant and the flanking bases, as well as the optimization of several phosphate contacts. The aggregate of biochemical data presented in this thesis leads us to propose a model where the A138T miscognate DNA binding ensemble (for a subset of miscognate sites) is partitioned more towards complexes that are on the path to the transition state than wild-type miscognate DNA binding ensembles; the mutant accomplishes this by forming 'specific-like' phosphate contacts to these sites, which in turn stabilize the DNA distortion that is critical for efficient cleavage. Given the proximity of amino acid 138 to the residues which make contacts to the DNA phosphates in the specific complex, we hypothesize that the A138T mutation has an effect on the structure and/or dynamics of the "arm" protein segment (contains residues contacting the DNA backbone) such that it is more adaptable, resulting in formation of functional phosphate contacts to a broader range of DNA substrates.

The interaction of light with matter is of fundamental importance, and is the mechanism that governs how photons are generated and used. Control over this interaction can be achieved by using an optical cavity, engineering the properties of the material to create a photonic crystal, or by using plasmonic devices to conne the electromagnetic field locally. Emitters can be coupled to devices with high-quality light confinement to enhance the emission rates, creating brighter light sources for detection, for contaminant sensing or for quantum information technologies. Such technologies require bright and pure-single photon sources: a solid-state platform can meet these requirements whilst also being compatible with well-developed fabrication technologies and long-term stability. We report metallic nanorings fabricated around selected solid-state quantum dots, to enhance vertical emission for collection by free-space optics, using a nanometre-accurate positioning technique. Enhancements of a single emission line as high as 25 are recorded thanks to a broadband lensing effect. Such metallic nanorings can be combined with deterministically-deposited super-solid immersion lenses, to provide further enhancement, 10, to that of the nanoring { creating photon sources with up to 1MHz emission rates. The light-matter interaction can be modified by using photonic crystals. The performance of photonic crystal devices in the visible light regime is hampered by unavoidable fabrication imperfections, which affect devices on the lengthscales required for visible light operation. An alternative to such highly engineered devices is to use the fabrication disorder as a resource. Anderson localization is demonstrated using photonic crystal waveguides, and directly imaged for the first time in the visible on a nanophotonic chip. Spectral analysis shows Q-factors approaching 10,000, exceeding highly engineered devices for the first time. Optical sensing is demonstrated by making use of high-quality resonances from photonic crystals to perform sensing: a contaminant was introduced, and a resonance was shown to red-shift over 100 its line-width in response. The resonances are also sensitive to temperature, shifting about 2nm under a 290K change. An alternative localization mechanism is to use aperiodic structures, where a quasi-random pattern is procedurally generated. We have fabricated and optically characterised aperiodic structures, showing efficient light confinement in a 2D system.

A new class of lasers has recently emerged: biolasers. Biolasers are sources of stimulated emission suitable for tissue integration. They have begun to captivate the scientific community due to their potential to harness the amplifying power of stimulated emission for biosensing, cell tracking and tagging. The main goal of this thesis is to fabricate a biocompatible random laser, study its properties and demonstrate a proof of sensing. Random lasers rely on multiple scattering to provide optical trapping necessary for lasing. Nanostructuring promotes light confinement while the gain provides light amplification. The lasing action is independent of the overall shape and in-stead relies on the device’s internal porosity; it can therefore easily adapt to biological media with the ability to withstand stretching, wetness and heat. Firstly we address how engineering light-matter interaction with a single biocompatible material can lead to increased light scattering and increased opaqueness. We demonstrate simple fabrication techniques which allow the nanostructuring of biomaterials. The gain is studied by doping biomaterials such as proteins, polysaccharides and silica and fabricating doped solid state whispering gallery mode laser. The threshold is evaluated as well as the biocompatibility. The different fabrication techniques are complementary and yield the first silk biocompatible random laser, where the technique is not limited to this material. The lasing output is studied in detail for different gain molecules. Miniature spherical random lasers are made and studied in terms of the threshold size dependence. In simultaneous we take on a different approach and develop macroporous lasers using freeze-drying. As proof of principle we explore lasing as a mechanism for enhanced sensing, demonstrating that biorandom lasing acts as a nonlinear sensor that switches off lasing at high pH values, with a sensitivity ~100 times larger than its fluorescent counterpart. Additionally we build a theoretical framework which clarifies the sensing mechanism. The combination of a natural biopolymer and random lasing offers the oppor-tunity for integration of a laser sensor within living tissue, opening a new path at the interface between nanophotonics and medicine.

China is a modernizing agrarian society. As such, it pursued certain goals within an environment of resource scarcity. This research analyzes Chinese public policies for agrarian development within a framework of political economy. It begins with the hypothesis that resources, (material output, infrastructure, political legitimacy, etc.), can be politicized such that their values becane relative within a single costs/benefits framework". The exchange of resources between the regime and social sectors is reflected in public policy. The hypothesis is tested in a longitudinal case study of Chinese agrarian policies, extending from 1949 through 1964. During this period, Chinese decision makers focused on the agrarian sector and its role in the attainment of national goals. Therefore, within this time frame, the dynamics of resource creation, investment and exchange can be traced through the agrarian policy matrix to demonstrate the utility of this model. When resources are politicized they are brought into a "political market" where they become available for use. The first step toward the creation of this market is the politicization of existing resources. In China, this was effected by land reform and by the creation of infrastructure to penetrate and control the rural marketplace itself. As the political market expands, it also becomes more complex, generating greater demands for all types of resources. Chinese collectivization policies reorganized the internal resources of the agrarian sector -- land and labor -- in order to increase their productivity. However, internal resources proved insufficient for agrarian modernization and the collective infrastructure proved costly in terms of political and social resources. Therefore, communization was adopted as an alternative pattern of resource use and investment. The size, functional scope and structural efficiency of the commune were designed to maximize the use-value of internal agrarian resources. But its high risk component was the introduction of a division of labor into the rural environment. Differentiation and specialization in the production process would destroy traditional socio-economic units which were functionally integrated. In terms of the political market, these policies would secure and control the availability and value of agrarian resources. The immediate costs of rapid, qualitative change quickly surpassed long-term benefits. During' the crisis period of 1959-1961, the division of labor was withdrawn, along with all but the, administrative functions of the commune. However, political inflation, followed by a drastic deflation, resulted in the collapse of the political market. Still seeking workable alternatives for resource utilization, the regime combined agrarian "self-reliance" with selective technological and capital investments. This strategy promised differentiated development within the sector and a perpetual rural/urban dichotomy. As such, it was opposed by the "Maoist" faction. Reconsolidation policies -- rebuilding the political market -- became infused with an ideological debate. Out of this struggle there emerged a workable and essentially Maoist approach -- the Dazhai model -- which created a supportive environment for long term agrarian development. The political economy of agrarian development energes when these policies are viewed within a single costs/benefits framework. Long range goals were held constant by ideology while intermediate aims were pursued by evaluating and exchanging resources, and by choosing among options for resource utilization and investment. The agrarian policy matrix between 1949 and 1964 thus emerges as complex and non-linear. But it is developmental, in that the aggregate level of resource availability and political productivity was increased.

As the acknowledged leader of Russian formalism, Viktor Sklovskij is most studied as a literary critic and scholar. Sklovskij's form of literary criticism and theorizing, however, is a unique blend of the theoretic and the poetic. In Sklovskij's texts, poetry enters metapoetic terminology (i.e., his tropes, figures, and key words perform theoretic, critical functions), a fact that has received insufficient scholarly attention. This dissertation is a descriptive and interpretive study of selected metapoetic imagery in Sklovskij's texts published in the years 1914 to 1933. It traces his use of reiterative tropes, figures, and key words that constitute the metapoetic images in his rhetorical dictionary and attempts to uncover their origins as well as locate their meanings. The major metapoetic tropes the dissertation treats are: (1) the stone; (2) the bomb: unity, destruction, and fragments; (3) ploughing and new growth; (4) the nomad and the horse; (5) the machine and work; (6) music, noise, and deafness; (7) the grill, the cage, and the net; and (8) love. Other related but minor images also figure in the discussion. My analysis makes wide use of Kiril Taranovsky's concept of quotation-introduced subtexts of such imagery, and the dissertation seeks to contribute to the work in the field of semiotics that studies critical (synchronic) and historical (diachronic) terminology in literary scholarship. This study shows the value of and need for examining Sklovskij's (meta)poetics as a means to better understanding his theories of literature. It also demonstrates the large role that subtext plays in the construction of Sklovskij's texts, most particularly the tremendous influence that the writings of contemporary Russian poets, particularly non-futurists such as Andrej Belyj, Aleksandr Blok, and Osip Mandel'stam, had on Sklovskij's own ideas about verbal art.

It is the purpose of this study to determine the extent of the need for an organized training program on the graduate level for vocational rehabilitation counselors. Specifically, the study proposes to discover: l, The nature of professional training which a selected group of vocational rehabilitation counselors in the New England area would like to develop through further education. 2, The extent of their personal interest in securing such training through university study or extension courses.

Light-matter interaction, which is at the heart of the science and technology of light, can be controlled and designed in materials structured at the nanometre length-scale, for enhanced light emission and absorption, and down to the single photon level. The aim of this thesis is to investigate complex photonic geometries, i.e. systems where the collective interaction of a large number of constituents denes the optical properties with emergent phenomena beyond the sum of the response of the individual constituents. In particular, a central topic is the emission of light from sources located in dielectric and plasmonic networks with dierent degree of disorder and correlation. Experimental and theoretical evidence of coupling of single photons to propagating modes in nano-waveguides, emission enhancement in plasmonic structures, and collective emission in disordered lasing systems are presented. Large coupling of individual quantum dots embedded in free-standing sub-wavelength waveguides is experimentally demonstrated. These waveguides are fabricated by electrospinning, a scalable technique suitable for the realisation of large interconnected systems. Light emission enhancement is investigated in plasmonic self-assembled systems and lithographic structures, which build on the framework of optical antennas and allows isolating local and global contributions to the local density of states around a topological percolation phase transition. One of the most important cooperative eects between multiple emitters is lasing. Random lasing is investigated numerically and experimentally in diusive systems with particular attention to the spectral properties of the emission and its relation with the physical and chemical parameters of the surrounding environment, which can be exploit to tune the lasing emission, thus providing a novel sensing scheme. These results provide the building blocks to construct a photonic network of emitters coupled by simple optical links.