This thesis addresses the problem of deriving a coherent and accurate localization of moving objects from partial visual information when data are generated by cameras placed in di erent view angles with respect to the scene. The framework is built around applications of scene monitoring with multiple cameras. Firstly, we demonstrate how a geometric-based solution exploits the relationships between corresponding feature points across views and improves accuracy in object location. Then, we improve the estimation of objects location with geometric transformations that account for lens distortions. Additionally, we study the integration of the partial visual information generated by each individual sensor and their combination into one single frame of observation that considers object association and data fusion. Our approach is fully image-based, only relies on 2D constructs and does not require any complex computation in 3D space. We exploit the continuity and coherence in objects' motion when crossing cameras' elds of view. Additionally, we work under the assumption of planar ground plane and wide baseline (i.e. cameras' viewpoints are far apart). The main contributions are: i) the development of a framework for distributed visual sensing that accounts for inaccuracies in the geometry of multiple views; ii) the reduction of trajectory mapping errors using a statistical-based homography estimation; iii) the integration of a polynomial method for correcting inaccuracies caused by the cameras' lens distortion; iv) a global trajectory reconstruction algorithm that associates and integrates fragments of trajectories generated by each camera.

This thesis presents research on several independent systems that when combined together can generate an automatic sound mix out of an unknown set of multi‐channel inputs. The research explores the possibility of reproducing the mixing decisions of a skilled audio engineer with minimal or no human interaction. The research is restricted to non‐time varying mixes for large room acoustics. This research has applications in dynamic sound music concerts, remote mixing, recording and postproduction as well as live mixing for interactive scenes. Currently, automated mixers are capable of saving a set of static mix scenes that can be loaded for later use, but they lack the ability to adapt to a different room or to a different set of inputs. In other words, they lack the ability to automatically make mixing decisions. The automatic mixer research depicted here distinguishes between the engineering mixing and the subjective mixing contributions. This research aims to automate the technical tasks related to audio mixing while freeing the audio engineer to perform the fine‐tuning involved in generating an aesthetically‐pleasing sound mix. Although the system mainly deals with the technical constraints involved in generating an audio mix, the developed system takes advantage of common practices performed by sound engineers whenever possible. The system also makes use of inter‐dependent channel information for controlling signal processing tasks while aiming to maintain system stability at all times. A working implementation of the system is described and subjective evaluation between a human mix and the automatic mix is used to measure the success of the automatic mixing tools.

The difficulty of semantic multimedia analysis can be attributed to the extended diversity in form and appearance exhibited by the majority of semantic concepts and the difficulty to express them using a finite number of patterns. In meeting this challenge there has been a scientific debate on whether the problem should be addressed from the perspective of using overwhelming amounts of training data to capture all possible instantiations of a concept, or from the perspective of using explicit knowledge about the concepts’ relations to infer their presence. In this thesis we address three problems of pattern recognition and propose solutions that combine the knowledge extracted implicitly from training data with the knowledge provided explicitly in structured form. First, we propose a BNs modeling approach that defines a conceptual space where both domain related evi- dence and evidence derived from content analysis can be jointly considered to support or disprove a hypothesis. The use of this space leads to sig- nificant gains in performance compared to analysis methods that can not handle combined knowledge. Then, we present an unsupervised method that exploits the collective nature of social media to automatically obtain large amounts of annotated image regions. By proving that the quality of the obtained samples can be almost as good as manually annotated images when working with large datasets, we significantly contribute towards scal- able object detection. Finally, we introduce a method that treats images, visual features and tags as the three observable variables of an aspect model and extracts a set of latent topics that incorporates the semantics of both visual and tag information space. By showing that the cross-modal depen- dencies of tagged images can be exploited to increase the semantic capacity of the resulting space, we advocate the use of all existing information facets in the semantic analysis of social media.

The task of automatic music transcription has been studied for several decades and is regarded as an enabling technology for a multitude of applications such as music retrieval and discovery, intelligent music processing and large-scale musicological analyses. It refers to the process of identifying the musical content of a performance and representing it in a symbolic format. Despite its long research history, fully automatic music transcription systems are still error prone and often fail when more complex polyphonic music is analysed. This gives rise to the question in what ways human knowledge can be incorporated in the transcription process. This thesis investigates ways to involve a human user in the transcription process. More specifically, it is investigated how user input can be employed to derive timbre models for the instruments in a music recording, which are employed to obtain instrument-specific (parts-based) transcriptions. A first investigation studies different types of user input in order to derive instrument models by means of a non-negative matrix factorisation framework. The transcription accuracy of the different models is evaluated and a method is proposed that refines the models by allowing each pitch of each instrument to be represented by multiple basis functions. A second study aims at limiting the amount of user input to make the method more applicable in practice. Different methods are considered to estimate missing non-negative basis functions when only a subset of basis functions can be extracted based on the user information. A method is proposed to track the pitches of individual instruments over time by means of a Viterbi framework in which the states at each time frame contain several candidate instrument-pitch combinations. A transition probability is employed that combines three different criteria: the frame-wise reconstruction error of each combination, a pitch continuity measure that favours similar pitches in consecutive frames, and an explicit activity model for each instrument. The method is shown to outperform other state-of-the-art multi-instrument tracking methods. Finally, the extraction of instrument models that include phase information is investigated as a step towards complex matrix decomposition. The phase relations between the partials of harmonic sounds are explored as a time-invariant property that can be employed to form complex-valued basis functions. The application of the model for a user-assisted transcription task is illustrated with a saxophone example.

There are two components of communication that provide a natural index of mutual-understanding in dialogue. The first is Repair; the ways in which people detect and deal with problems with understanding. The second is Ellipsis/Anaphora; the use of expressions that depend directly on the accessibility of the local context for their interpretation. This thesis explores the use of these two phenomena in systematic comparative analyses of human-human dialogue under different task and media conditions. In order to do this it is necessary to a) develop reliable, valid protocols for coding the different Repair and Ellipsis/Anaphora phenomena b) establish their baseline patterns of distribution in conversation and c) model their basic statistical inter-relationships and their predictive value. Two new protocols for coding Repair and Ellipsis/Anaphora phenomena are presented and applied to two dialogue corpora, one of ordinary 'everyday' conversations and one of task-oriented dialogues. These data illustrate that there are significant differences in how understanding is created and negotiated across conditions. Repair is shown to be a ubiquitous feature in all dialogue. The goals of the speaker directly affect the type of Repair used. Giving instructions leads to a higher rate of self-editing; following instructions increases corrections and requests for clarification. Medium and familiarity also influence Repair; when eye contact is not possible there are a greater number of repeats and clarifications. Anaphora are used less frequently in task-oriented dialogue whereas types of Ellipsis increase. The use of Elliptical phrases that check, confirm or acknowledge is higher when there is no eye contact. Familiar pairs use more elliptical expressions, especially endophora and elliptical questions. Following instructions leads to greater use of elliptical (non-sentential) phrases. Medium, task and social norms all have a measureable effect on the components of dialogue that underpin mutual-understanding.

New services are emerging rapidly within the world of telecommunications. Charging strategies that were appropriate for individual transfer capabilities are no longer appropriate for an integrated broadband communications network. There is currently a range of technologies (such as cable television, telephony and narrow band ISDN) for the different services in use and a limited number of charging schemes are applicable for each of the underlying technologies irrespective of the services used over it. Difficulties arise when a wide range of services has to be supported on the same integrated technology such as asynchronous transfer mode (ATM); in such cases the type of service in use and the impact it has on the network becomes much more important. The subject of this thesis, therefore, is the charging strategies for integrated broadband communications networks. That is, the identification of the requirements associated with ATM charging schemes and the proposal of a new approach to charging for ATM called the “quality of service based charging scheme”. Charging for ATM is influenced by three important components: the type and content of a service being offered; the type of customer using the services; and the traffic characteristics belonging to the application supporting the services. The first two issues will largely be dependent on the business and regulatory requirements of the operators. The last item, and an essential one for ATM, is the bridge between technology and business; how are the resources used by a service quantified? Charging that is based on resource usage at the network level was the prime focus of the research reported here. With the proposed charging scheme, a distinction is first made between the four different ATM transfer capabilities that will support various services and the different quality of service requirements that may be applicable to each of them. Then, resources are distributed among buffers set-up to support the combination of these transfer capabilities and quality of services. The buffers are dimensioned according to the M/D/1/K and the ND/D/1 queuing analysis to determine the buffer efficiency and quality of service requirements. This dimensioning provides the basis for fixing the price per unit of resource and time. The actual resource used by a connection is based on the volume of cells transmitted or peak cell rate allocation in combination with traffic shapers if appropriate. Shapers are also dimensioned using the quality of service parameters. Since the buffer 4 efficiency is dependent on the quality of service requirements, users (customers) of ATM networks buy quality of service. The actual price of a connection is further subjected to a number of transformations based on the size of the resource purchased, the time of the day at which a connection is made, and the geographical locality of the destination switch. It is demonstrated that the proposed charging scheme meets all the requirements of customers and of network operators. In addition the result of the comparison of the new scheme with a number of existing, prominent, ATM charging schemes is presented, showing that the performance of the proposed scheme is better in terms of meeting the expectations of both the customers and the network operators.

No description available.

During this PhD study, several finite-difference time-domain (FDTD) methods were developed to numerically investigate coordinate transformation based metamaterial devices. A novel radially-dependent dispersive FDTD algorithm was proposed and applied to simulate electromagnetic cloaking structures. The proposed method can ac- curately model both lossless and lossy cloaks with ideal or reduced parameters. It was demonstrated that perfect “invisibility” from electromagnetic cloaks is only available for lossless metamaterials and within an extremely narrow frequency band. With a few modifications the method is able to simulate general media, such as concentrators and rotation coatings, which are produced by means of coordinate transformations techniques. The limitations of all these devices were thoroughly studied and explo- red. Finally, more useful cloaking structures were proposed, which can operate over a broad frequency spectrum. Several ways to control and manipulate the loss in the electromagnetic cloak ba- sed on transformation electromagnetics were examined. It was found that, by utili- sing inherent electric and magnetic losses of metamaterials, as well as additional lossy materials, perfect wave absorption can be achieved. These new devices demonstrate super-absorptivity over a moderate wideband range, suitable both for microwave and optical applications. Furthermore, a parallel three-dimensional dispersive FDTD method was introdu- ced to model a plasmonic nanolens. The device has its potential in subwavelength imaging at optical frequencies. The finiteness of such a nano-device and its impact on the system dynamic behaviour was numerically exploited. Lastly, a parallel FDTD method was also used to model another interesting coordinate transformation based device, an optical black hole, which can be characterised as an omnidirectional broad- band absorber.

This thesis focuses on applying adaptive system level cooperative coverage to wireless communication networks and proposes an Adaptive Coverage System (ACS) for flexible and resilient radio coverage to handle time-varying geographic traffic distribution, providing better than conventional radio resource allocation results. The ACS utilises antenna arrays called Semi-Smart Antennas to shape cell coverage to alleviate cell overloading and balance loads as equally as possible among cells. Previously work in this field used a Bubble Oscillation Algorithm (BOA) to solve the load balancing problem in the ACS. Modification is made to the core algorithm to enhance its performance for its applications in WCDMA networks, to handle the impact of complex radio propagation environment. The work presented in this thesis then extends into tackling the problem from an optimisation problem point of view. It firstly examines the dynamics of user assignment in a network level and based on these findings, the possibility and approaches towards finding optimal coverage shapes for a mobile network are then presented, including problem formulation, algorithm design, and simulation results. This thesis investigates a system wide cooperative coverage approach to explore the potentials of adaptive coverage systems, has achieved sub-optimal radio resource allocation results, and provides an insight into general network-level radio resource allocation problems and their solutions.

This dissertion explores the relation between two formalisms and one algebraic framework for concurrency. Session Types and Concurrent Separation Logic are formalisms that support independent reasoning about concurrent processes, and our motivating question is whether their modularity springs from the same source despite the distance between their models. We first translate a small language we call Baby Session Types (BST), into a ‘basic’ version of Concurrent Separation Logic (BCSL), and we show that the translation is sound. We then describe a model for Separation Logic (SL) based on Actions, which exhibits some of the structure of a Concurrent Kleene Algebra, an algebra where operators for parallel and sequential composition are linked by a version of the exchange law from category theory. The model connects the algebraic notions to locality concepts that underlie Separation Logic. We then move on to provide a more general construction of an algebra model of BCSL, which can be built from (Baby) Session Types. Thus, we end up with a model that brings together concepts from all of Session Types, Separation Logic, and Concurrent Kleene Algebra. Thus, the model links diverse models of concurrency. In addition to this it suggests alterations of the algebraic axioms as well as the foundational models underlying Separation Logic. It is hoped that, apart from these specific results, this dissertation can in some modest way contribute to unification in concurrency theory, a theory (or theories) based presently on diverse models.