Data Fusion based on Game Theory for Speaker Diarization

A novel algorithm based on bimatrix game theory has been developed to improve the accuracy and reliability of a speaker diarization system. This algorithm fuses the output data of two open-source speaker diarization programs, LIUM and SHoUT, taking advantage of the best properties of each one. The performance of this new system has been tested by means of audio streams from several movies. From preliminary results on fragments of five movies, improvements of 63% in false alarms and missed speech mistakes have been achieved with respect to LIUM and SHoUT systems working alone. Moreover, we also improve in a 20% the number of recognized speakers, getting close to the real number of speakers in the audio stream

Non uniform embedding based on relevance analysis with reduced computational complexity: application to the detection of pathologies from biosignal recordings

Nonlinear analysis tools for studying and characterizing the dynamics of physiological signals have gained popularity, mainly because tracking sudden alterations of the inherent complexity of biological processes might be an indicator of altered physiological states. Typically, in order to perform an analysis with such tools, the physiological variables that describe the biological process under study are used to reconstruct the underlying dynamics of the biological processes. For that goal, a procedure called time-delay or uniform embedding is usually employed. Nonetheless, there is evidence of its inability for dealing with non-stationary signals, as those recorded from many physiological processes. To handle with such a drawback, this paper evaluates the utility of non-conventional time series reconstruction procedures based on non uniform embedding, applying them to automatic pattern recognition tasks. The paper compares a state of the art non uniform approach with a novel scheme which fuses embedding and feature selection at once, searching for better reconstructions of the dynamics of the system. Moreover, results are also compared with two classic uniform embedding techniques. Thus, the goal is comparing uniform and non uniform reconstruction techniques, including the one proposed in this work, for pattern recognition in biomedical signal processing tasks. Once the state space is reconstructed, the scheme followed characterizes with three classic nonlinear dynamic features (Largest Lyapunov Exponent, Correlation Dimension and Recurrence Period Density Entropy), while classification is carried out by means of a simple k-nn classifier. In order to test its generalization capabilities, the approach was tested with three different physiological databases (Speech Pathologies, Epilepsy and Heart Murmurs). In terms of the accuracy obtained to automatically detect the presence of pathologies, and for the three types of biosignals analyzed, the non uniform techniques used in this work lightly outperformed the results obtained using the uniform methods, suggesting their usefulness to characterize non-stationary biomedical signals in pattern recognition applications. On the other hand, in view of the results obtained and its low computational load, the proposed technique suggests its applicability for the applications under study.

Advanced background modeling with RGB-D sensors through classifiers combination and inter-frame foreground prediction

An innovative background modeling technique that is able to accurately segment foreground regions in RGB-D imagery (RGB plus depth) has been presented in this paper. The technique is based on a Bayesian framework that efficiently fuses different sources of information to segment the foreground. In particular, the final segmentation is obtained by considering a prediction of the foreground regions, carried out by a novel Bayesian Network with a depth-based dynamic model, and, by considering two independent depth and color-based mixture of Gaussians background models. The efficient Bayesian combination of all these data reduces the noise and uncertainties introduced by the color and depth features and the corresponding models. As a result, more compact segmentations, and refined foreground object silhouettes are obtained. Experimental results with different databases suggest that the proposed technique outperforms existing state-of-the-art algorithms.

A Ground-Truth Video Dataset for the Development and Evaluation of Vision-based Sense-and-Avoid systems

The importance of vision-based systems for Sense-and-Avoid is increasing nowadays as remotely piloted and autonomous UAVs become part of the non-segregated airspace. The development and evaluation of these systems demand flight scenario images which are expensive and risky to obtain. Currently Augmented Reality techniques allow the compositing of real flight scenario images with 3D aircraft models to produce useful realistic images for system development and benchmarking purposes at a much lower cost and risk. With the techniques presented in this paper, 3D aircraft models are positioned firstly in a simulated 3D scene with controlled illumination and rendering parameters. Realistic simulated images are then obtained using an image processing algorithm which fuses the images obtained from the 3D scene with images from real UAV flights taking into account on board camera vibrations. Since the intruder and camera poses are user-defined, ground truth data is available. These ground truth annotations allow to develop and quantitatively evaluate aircraft detection and tracking algorithms. This paper presents the software developed to create a public dataset of 24 videos together with their annotations and some tracking application results.

Selectively filled photonic liquid crystal fibers = Fibras de cristal fotónico selectivamente llenas con cristal líquido

El presente trabajo de Tesis se ha centrado en el diseño, fabricación y caracterización de dispositivos basados en fibras de cristal fotónico infiltrado selectivamente con cristales líquidos, polímeros y una mezcla de ambos. Todos los dispositivos son sintonizables, y su área de aplicación se centra en comunicaciones ópticas y sensores. La manipulación y fusionado de fibras fotónicas, el llenado selectivo de determinadas cavidades y la alineación recíproca de fibras mantenedoras de polarización son tareas muy específicas y delicadas para las que se requieren protocolos muy estrictos. Previo a la fabricación de dispositivos ha sido necesaria por tanto una tarea de sistematización y creación de protocolos de fabricación. Una vez establecidos se ha procedido a la fabricación y caracterización de dispositivos. Los dispositivos fabricados se enumeran a continuación para posteriormente detallar una a una las singularidades de cada uno. • Interferómetros intermodales hechos a partir de una porción de fibra fotónica soldada entre dos fibras estándar, bien monomodo o PANDA (mantenedora de polarización). Estos interferómetros han sido sumergidos o bien llenados selectivamente con cristales líquidos para así sintonizar la señal interferométrica guiada a través de la fibra. • Infiltración de fibras fotónicas con cristales líquidos colestéricos con especial énfasis en la fase azul (blue phase) de estos materiales. Las moléculas de cristal líquido se autoalinean en volumen por lo que la infiltración de fibras fotónicas con estos cristales líquidos es muy interesante, pues es conocida la dificultad de alinear apropiadamente cristales líquidos dentro de cavidades micrométricas de las fibras fotónicas. • Grabación de redes holográficas de forma selectiva en las cavidades de una fibra fotónica. Estas redes holográficas, llamadas POLICRYPS (POlymer-LIquid CRYstal-Polymer Slices), son redes fabricadas a base de franjas de polímero y cristal líquido alineado perpendicularmente a dichas franjas. Las franjas son a su vez perpendiculares al eje de la fibra como lo puede ser una red de Bragg convencional. El cristal líquido, al estar alineado perpendicularmente a dichos franjas y paralelo al eje de la fibra, se puede conmutar aplicando un campo eléctrico externo, modificando así el índice efectivo de la red. Se puede fabricar por lo tanto una red de Bragg sintonizable en fibra, muy útil en comunicaciones ópticas. • Llenado selectivo de fibras fotónicas con polidimetilsiloxano (PDMS), un polímero de tipo silicona. Si se realiza un llenado selectivo asimétrico se puede inducir birrefringencia en la fibra. El índice de refracción del PDMS tiene una fuerte dependencia térmica, por lo que se puede sintonizar la birrefringencia de la fibra. • Estudio teórico de llenado selectivo de fibras fotónicas con PDMS dopado con nanopartículas de plata de 5, 40 y 80 nm. Estas nanopartículas poseen un pico de absorción en torno a los 450 nm debido a resonancias superficiales localizadas de plasmones (LSPR). La resonancia del plasmon tiene una fuerte dependencia con el índice de refracción del material colindante, y al ser éste PDMS, la variación de índice de refracción se ve amplificada, obteniendo una absorción sintonizable. Se ha propuesto la fabricación de polarizadores sintonizables usando esta técnica. Como ya se ha dicho, previamente a la fabricación ha sido necesaria la protocolización de diversos procedimientos de fabricación de alta complejidad, así como protocolizar el proceso de toma de medidas para optimizar los resultados. Los procedimientos que han requerido la formulación de protocolos específicos han sido los siguientes: • Llenado selectivo de cavidades en una fibra fotónica. Dichas fibras tienen generalmente un diámetro externo de 125 μm, y sus cavidades son de entre 5 y 10 μm de diámetro. Se han desarrollado tres técnicas diferentes para el llenado/bloqueado selectivo, pudiéndose combinar varios protocolos para la optimización del proceso. Las técnicas son las siguientes: o Llenado y bloqueado con un prepolímero. Dicho prepolímero, también llamado adhesivo óptico, está inicialmente en estado líquido y posee una cierta viscosidad. Las cavidades de la fibra fotónica que se desea llenar o bloquear poseen un diámetro diferente al resto, por lo que en el proceso de llenado aparecen dos frentes de llenado dependientes de su diámetro. A mayor diámetro, mayor velocidad de llenado. Polimerizando cuando existe dicha diferencia en los frentes se puede cortar por medio, obteniendo así una fibra parcialmente bloqueada. o Colapsamiento de las cavidades de menor diámetro mediante aplicación de calor. El calor producido por un arco voltaico de una soldadora de fibra estándar fusiona el material exterior de la fibra produciendo el colapsamiento de las cavidades de menor diámetro. En esta técnica también es necesaria una diferencia de diámetros en las cavidades de la fibra. o Bloqueo una a una de las cavidades de la fibra fotónica con adhesivo óptico. Este procedimiento es muy laborioso y requiere mucha precisión. Con este sistema se pueden bloquear las cavidades deseadas de una fibra sin importar su diámetro. • Alineación de una fuente de luz linealmente polarizada con una fibra mantenedora de polarización ya sea PANDA o fotónica. Así mismo también se han alineado entre sí fibras mantenedoras de polarización, para que sus ejes rápidos se fusionen paralelos y así el estado de polarización de la luz guiada se mantenga. • Sistematización de toma de medidas para caracterizar los interferómetros modales. Éstos son altamente sensibles a diversas variables por lo que el proceso de medida es complejo. Se deben aislar variables de forma estrictamente controlada. Aunque todos los dispositivos tienen en común el llenado selectivo de cavidades en una fibra fotónica cada dispositivo tiene sus peculiaridades, que van a ser explicadas a continuación. ABSTRACT The present Thesis has been centered in the design, fabrication and characterization of devices based on photonic crystal fibers selectively filled with liquid crystals, polymers and a mixture of both. All devices are tunable and their work field is optical communications and sensing The handling and splicing of photonic crystal fibers, the selective filling of their holes and the aligning of polarization maintaining fibers are very specific and delicate tasks for which very strict protocols are required. Before the fabrication of devices has therefore been necessary task systematization and creation of manufacturing protocols. Once established we have proceeded to the fabrication and characterization of devices. The fabricated devices are listed below and their peculiarities are detailed one by one: • Intermodal interferometers made with a portion of photonic crystal fiber spliced between two optical communication fiber pigtails, either single mode or PANDA (polarization-maintaining) fiber. These interferometers have been submerged or selectively filled with liquid crystals to tune the interferometric guided signal. • Infiltration of photonic fibers with cholesteric liquid crystals with special emphasis on their blue phase (blue phase). The liquid crystal molecules are self-aligning in volume so the infiltration of photonic fibers with these liquid crystals is very interesting. It is notoriously difficult to properly align liquid crystals within micron cavities such as photonic fibers. • Selectively recording of holographic gratings in the holes of photonic crystal fibers. These holographic gratings, called POLICRYPS (POlymer-LIquid CRYstal-Polymes Slices), are based on walls made of polymer and liquid crystal aligned perpendicular to them. These walls are perpendicular to the axis of the fiber as it can be a conventional Bragg grating. The liquid crystal is aligned perpendicular to the walls and parallel to the fiber axis, and can be switched by applying an external electric field and thus change the effective index of the grating. It is thus possible to manufacture a tunable Bragg grating fiber, useful in optical communications. •Asymmetrically selective filling of photonic crystal fibers with a silicone polymer like called polydimethylsiloxane (PDMS) to induce birefringence in the fiber. The refractive index of PDMS has temperature dependence, so that the birefringence of the fiber can be tuned. • Theoretical study of photonic crystal fibers selectively filled with PDMS doped with silver nanoparticles of 5, 40 and 80 nm. These nanoparticles have an absorption peak around 450 nm due to localized surface plasmon resonances (LSPR). Plasmon resonance has a strong dependence on the refractive index of the adjacent material, and as this is PDMS, the refractive index variation is amplified, obtaining a tunable absorption. Fabrication of tunable polarizers using this technique has been proposed. Before starting the fabrication, it has been necessary to optimize several very delicate procedures and different protocols have been designed. The most delicate procedures are as follows: • Selective filling of holes in a photonic crystal fiber. These fibers generally have an outer diameter of 125 μm, and their holes have a diameter around between 5 and 10 μm. It has been developed three different techniques for filling / selective blocking, and they can be combined for process optimization. The techniques are: o Filling and blocked with a prepolymer. This prepolymer also called optical adhesive is initially in liquid state and has a certain viscosity. The holes of the photonic crystal fiber that are desired to be filled or blocked should have a different diameter, so that in the filling process appear two different fronts depending on the hole diameter. The holes with larger diameter are filled faster. Then the adhesive is polymerized when there is such a difference on the front. A partially blocked fiber is obtained cutting between fronts. o Collapsing of holes of smaller diameter by application of heat. The heat produced by an arc of a standard fusion splicer fuses the outer fiber material producing the collapsing of the cavities of smaller diameter. In this technique also you need a difference of diameters in the fiber holes. o Blocking one by one the holes of photonic crystal fiber with optical adhesive. This procedure is very laborious and requires great precision. This system can block unwanted cavities regardless fiber diameter. • Aligning a linearly polarized light source with a polarization-maintaining fiber (either a PANDA fiber as a photonic crystal fiber). It is needed also an aligning between polarization-maintaining fibers, so that their fast axes parallel merge and that is state of polarization of light guided is maintained. • Systematization of taking measurements to characterize the modal interferometers. These are highly sensitive to several variables so the measurement process is very complicated. Variables must be fixed in a very controlled manner. Although all devices have the common characteristic of being selectively filled PCFs with some kind of material, each one has his own peculiarities, which are explained below.