A research report submitted to the Faculty of Engineering and the Built Environment, University of the Witwatersrand, Johannesburg, in partial fulfilment of the requirements for the degree of Master of Science in Engineering, 2017 / The images from the African articles dataset presented challenges to the Optical Character Recognition (OCR) tool. Despite successful binerisation in the Image Processing step of the pipeline, noise remained in the foreground of the images. This noise caused the OCR tool to misinterpret the text from the images and thus needed removal from the foreground. The technique involved the application of the Maximally Stable Extremal Region (MSER) algorithm, borrowed from Scene-Text Detection, and supervised machine learning classifiers. The algorithm creates regions from the foreground elements. Regions are classifiable into noise and characters based on the characteristics of their shapes. Classifiers were trained to recognise noise and characters. The technique is useful for a researcher wanting to process and analyse the large dataset. They could semi-automate the foreground noise-removal process using this technique. This would allow for better quality OCR output, for use in the Text Analysis step of the pipeline. Better OCR quality means less compromises would be required at the Text Analysis step. These concessions can lead to false results when searching noisy text. Fewer compromises means simpler, less error-prone analysis and more trustworthy results. The technique was tested against specifically selected images from the dataset which exhibited noise. It involved a number of steps. Training regions were selected and manually classified. After training and running many classifiers, the highest performing classifier was selected. The classifier categorised regions from all images. New images were created by removing noise regions from the original images. To discover whether an improvement in the OCR output was achieved, a text comparison was conducted. OCR text was generated from both the original and processed images. The two outputs of each image were compared for similarity against the test text. The test text was a manually created version of the expected OCR output per image. The similarity test for both original and processed images produced a score. A change in the similarity score indicated whether the technique had successfully removed noise or not. The test results showed that blotches in the foreground could be removed, and OCR output improved. Bleed-through and page fold noise was not removable. For images affected by noise blotches, this technique can be applied and hence less concessions will be needed when processing the text generated from those images. / CK2018

Noise pollution from road traffic is one of the greatest environmental issues in modern day, and the social cost for road traffic noise was estimated to over 16 billion SEK per year in Sweden in2014. Passive or active control methods can be used to reduce the noise. Active control methods or active noise control is more suitable for attenuating noise in lower frequencies. Active noise control reduces noise by eliminating the noise with a secondary source. There are different control strategies to construct an active noise control system, where the update of the secondary sourceis controlled by an algorithm. There are several different algorithms that are possible to use, and one option is to use a Feedforward Filtered-X Least-Mean-Square (FXLMS) algorithm. It uses control positions where the noise is meant to be reduced and reference signals that measure the noise upstream prior the secondary source. FXLMS also uses a model of the secondary source path to the control position in order to ensure convergence of the algorithm. Although the use of multiple reference signals increases the accuracy of the algorithm, it also increases the convergence time and the practical cost of such an installation. Unfortunately, it can require many reference signals to obtain a sufficient noise reduction when the unwanted noise source is complex and has multiple propagation paths.This study investigates the possibility of producing a new, reduced set of reference signals with a linear combination of the original reference signals that still contain the majority of information needed for suficient noise reduction. This new set of reference signals are sometimes called virtual reference signals. Three different methods of virtual reference signals are analysed; first a constant method using singular-value decomposition on the covariance of the reference signals, second another constant method using singular-value decomposition on the covariance of response estimate from each corresponding reference signal, third an adaptive algorithm updating the linear combination to adapt for incoming data. The different strategies are tested on road test measurements at three different constant speeds, 40km=h; 80km=h and 120km=h, and on data generated from a numerical vehicle model in COMSOL.The results from the analysis indicates that the virtual reference signals could sufficiently reproduce information from the original reference signals to obtain a similar noise reduction with fewer reference signals. However, the virtual reference signals with the adaptive algorithm could not manage to track a transient system where the signal amplitudes are varying over time. Further work is needed to analyse the limits and requirements to obtain virtual reference signals that can represent and track a system even for transient events.

Vacuum-assisted toilets make use of a large pressure difference between the ambient pressure and a vacuum tank to transport waste from the toilet bowl to the septic tank. This process requires 98% less water per flush making it an attractive product for transport vehicles such as airplanes, cruise ships, and trains. Unfortunately, the water savings come at the cost of high noise levels. This thesis investigates the acoustic characteristics of a vacuum-assisted toilet flush and several methods to reduce the radiated noise. Some methods include changing rinse parameters such as rinse pressure, rinse length, and rinse timing, adding structural damping of the bowl to reduce re-radiation, inserting a tube between the bowl and valve that utilizes a larger bend radius and longer tube length than what is currently installed, and modifying the valve. The most effective solution without requiring more water per flush was to insert a tube. The initial peak level was reduced by 16 dB and the steady-vacuum noise was reduced by 5 dB. Evidence of evanescent decay and reduced flow velocity as possible mechanisms for the noise reduction are presented and discussed. Rinse variations show a strong impact of the rinse-tube interaction on the noise reduction. In addition to these techniques, a modified flush plate opening and closing velocity profile is suggested which optimizes the sound generated by the opening and closing of the valve. Finally, a promising dual-valve solution that may take extra coordination of vacuum-assisted toilet manufacturers and airplane/cruise ship/train manufacturers is presented. By placing a secondary valve near the septic tank, the main noise from the valve is significantly reduced.

<p dir="ltr">As speeds and directions of the vehicle and wind change, the unsteady flow creates variations in wind noise, which can be referred to as non-stationary wind noise. To investigate people's perceptions of non-stationary wind noise, a method to simulate the non-stationary wind noise is needed. Previously, a method was developed that used stationary recordings taken at several wind speeds and directions to form functions that relate the 1/3 octave sound pressure level with wind speed and direction. These functions are used to create time-varying filters based on provided time histories of wind speed and direction. A reference wind noise measurement is then filtered to produce the sounds. To reduce the time cost of taking many stationary measurements, an improved method was investigated. At each yaw angle, one speed sweep wind tunnel measurement was used to estimate the relationship between sound pressure level and wind speed. Two partially correlated white noise signals were then filtered to simulate binaural sounds that had a similar coherence structure between the left and right ear sounds to that observed in binaural measurements in the vehicle. The accuracy of the simulations was validated by comparing wind noise simulations with wind tunnel and on-road vehicle interior noise measurements. For the on-road measurements, a noise decomposition method based on noise source measurements was used to estimate the road/tire noise, which was then added to the simulated wind noise to make it comparable with the measured on-road noise. The time-varying loudness, and power spectral densities of the simulated and measured sounds were found to be well consistent. Besides, a method to simulate the turbulent wind speed time histories, which can be used as inputs to the wind noise simulation method, was developed. The von Karman model predicts the spectra of wind turbulence by assuming it to be a stationary random process. White noise signals can then be filtered to simulate the stable variations of wind speeds. The discrete gusts, which are the transient events in wind speed time histories, were also simulated by using an 8-parameter piecewise function. Eventually, the non-stationary wind noise and the turbulent wind speed simulation method can be a powerful tool when investigating sound perceptions of vehicle interior wind noise.</p>

Conventional isolation models deal with massless isolators, which tend to overestimate the isolator performance because they neglect the internal resonances (IRs) due to the inertia of the isolator. Previous researches on the IR problem is not adequate because they only discussed this problem in terms of vibration based on single degree-of-freedom (SDOF) models. These studies did not reveal the importance of the IRs, especially from the perspective of the noise radiation. This dissertation is novel compared to previous studies in the following ways: (a) a three-DOF (3DOF) model, which better represents practical vibration systems, is employed to investigate the importance of the IRs; (b) the IR problem is studied considering both vibration and noise radiation; and (c) passive and hybrid control approaches using dynamic vibration absorbers (DVAs) to suppress the IRs are investigated and their potential demonstrated. The 3DOF analytical model consists of a rigid primary mass connected to a flexible foundation through three isolators. To include the IRs, the isolator is modeled as a continuous rod with longitudinal motion. The force transmissibility through each isolator and the radiated sound power of the foundation are two criteria used to show the effects and significance of the IRs on isolator performance. Passive and hybrid DVAs embedded in the isolator are investigated to suppress the IRs. In the passive approach, two DVAs are implemented and their parameters are selected so that the IRs can be effectively attenuated without significantly degrading the isolator performance at some other frequencies that are also of interest. It is demonstrated that the passive DVA enhanced isolator performs much better than the conventional isolator in the high frequency range where the IRs occur. The isolator performance is further enhanced by inserting an active force pair between the two passive DVA masses, forming the hybrid control approach. The effectiveness and the practical potential of the hybrid system are demonstrated using a feedforward control algorithm. It is shown that this hybrid control approach not only is able to maintain the performance of the passive approach, but also significantly improve the isolator performance at low frequencies. / Ph. D.

Trailing edge noise control is an important problem associated mainly with wind turbines. As turbulence in the air flows over a wind turbine blade, it impacts the trailing edge and scatters, producing noise. Traditional methods of noise control involve modifying the physical trailing edge, or the scattering efficiency. Recently, inspired by the downy covering of owl feathers, researchers developed treatments that can be applied to the trailing edge to significantly reduce trailing edge noise. It was hypothesized that the noise reduction was due to manipulating the incoming turbulence, rather than the physical trailing edge itself, representing a new method of noise control. However, only acoustic measurements were reported, meaning the associated flow physics were still unknown. This thesis describes a comprehensive wall jet experiment to measure the flow effects near the bio-inspired treatments, termed “finlets” and “rails,” and relate those flow effects to the noise reduction. This was done using far-field microphones, a single hot-wire probe, and surface pressure fluctuation microphones. The far-field noise results showed that each treatment successfully reduced the noise, by up to 7 dB in some cases. The surface pressure measurements showed that the spanwise coherence was slightly reduced when the treatments were applied to the trailing edge. The velocity measurements clearly established the presence of a shear layer near the top of the treatments. As a whole, the dataset led to the shear-sheltering hypothesis: the bio-inspired treatments are effective based on reducing the spanwise pressure correlation and by sheltering the trailing edge from turbulent structures with the shear layer they create. / Master of Science / This thesis describes a project aimed at developing a technology inspired by the silent flight of owls, with the end goal of using this technology to reduce the noise generated by wind turbines. Specifically, the phenomenon known as "trailing edge noise" is the primary source of wind turbine noise, and is the noise source of interest here. It occurs when air turbulence (which can be thought of as unsteady air fluctuations) crashes into the rear (trailing) edge of wind turbine blades, scattering and producing noise. Typically, methods of reducing this noise source involve changing the shape of the trailing edge; this may not always be practical for existing wind turbines. Recently, inspired by the downy covering of owl feathers, researchers developed treatments that can be applied directly to the trailing edge, significantly reducing trailing edge noise. This bio-inspired concept was verified with numerous acoustic measurements. Based on those measurements, researchers hypothesized that the noise reduction was achieved by manipulating the incoming turbulence before it scattered off the trailing edge, rather than by changing the existing wind turbine blade, representing a new method of trailing edge noise control. However, as only acoustic measurements (not flow measurements) were reported, the changes in turbulence could not be examined. With the above motivation in mind, this thesis describes a comprehensive wind tunnel experiment to measure the changes in the aerodynamics and turbulence near the bio-inspired treatments, and relate those changes to the reduction in trailing edge noise. This was done using a hot-wire probe to measure the aerodynamics, as well as microphones to measure the radiated noise and surface pressure fluctuations. As a whole, the experimental results led to the shear-sheltering hypothesis: the bio-inspired treatments are effective based on the creation of a shear layer (a thin region between areas with different air speeds) which shelters the trailing edge from some turbulence, as well as by de-correlating surface pressure fluctuations along the trailing edge.

<p dir="ltr">Acoustic holography is a powerful tool in the visualization of sound fields and sound sources. It provides engineers and researchers clear insights into sound fields as well as their sound sources. Some widely-used methods include Nearfield Acoustical Holography (NAH), Statistically Optimized Nearfield Acoustic Holography (SONAH) and the Equivalent Source Method (ESM). SONAH and ESM were developed specifically to tackle the intrinsic deficiency of the Fourier-based NAH which requires that the sound field fall to negligible levels at the edges of the measurement aperture, a requirement rarely met in practice. Besides the aforementioned methods, the Inverse Boundary Element Method (IBEM) can be used, given sufficient measurements and computational resources. As useful as they are in visualizing the sound field, none of these methods can provide direct guidance on potential design modifications of the observed structure in order to unequivocally reduce sound power radiation. Acoustic radiation mode analysis has previously been primarily associated with active noise control applications. Since the radiation modes radiate sound power independently, it is only necessary to modify the surface vibration patterns so that they do not couple well with the radiation modes in order to guarantee a reduction of the radiated sound power. Since the radiation modes are orthogonal and complete, they can be used as the basis functions through which the source surface vibration can be described. Therefore, an acoustic holography method based on the acoustic radiation modes will enable the sound power ranking of the modal components of the surface vibration pattern, and in turn, point out the component(s) which should be targeted in order to reduce the overall sound power. However, use of the acoustic radiation modes in the inverse procedure comes with a price: the detailed geometry of the object to be measured must be obtained, thus enabling the calculation of acoustic radiation modes and the modal pressures. But this is not an issue for original equipment manufacturers given that almost all prototypes are now designed with CAD, as is the case with the engine example to be described next.</p><p dir="ltr">In modern engine design, downsizing and reducing weight while still providing an increased amount of power has been a general trend in recent decades. Traditionally, an engine design with superior NVH performance usually comes with a heavier, thus sturdier structure. Therefore, modern engine design requires that NVH be considered in the very early design stage to avoid modifications of engine structures at the last minute, when very few changes can be made. NVH design optimization of engine components has become more practical due to the development of computer software and hardware. However, there is still a need for smarter algorithms to draw a direct relationship between the design and the radiated sound power. At the moment, techniques based on modal acoustic transfer vectors (MATVs) have gained popularity in design optimization for their good performance in sound pressure prediction. Since MATVs are derived based on structural modes, they are not independent with respect to radiated sound power. In contrast, as noted, acoustic radiation modes are an orthogonal set of velocity distributions on the structure’s surface that contribute to the radiated sound power independently. As a result, it is beneficial to describe structural vibration in terms of acoustic radiation modes in order to identify the velocity distributions that contribute the majority of the radiated sound power. Measures can then be taken to modify the identified vibration patterns to reduce their magnitudes, which will in turn result in an unequivocal reduction of the radiated sound power. A workflow of the structural optimization procedure is proposed in this dissertation.</p><p dir="ltr">While acoustic radiation modes have great efficiencies in describing radiated acoustic power, the computation of acoustic radiation modes can be time consuming. In the last chapter of this thesis, a novel way of calculating acoustic radiation modes is proposed, which differs from the traditional singular value decomposition of the power radiation resistance matrix, and which is more efficient than previously proposed procedures. </p><p><br></p>

An aeroacoustic study was conducted on a 26%-scale Boeing 777 main landing gear in the Virginia Tech (VT) Anechoic Stability Wind Tunnel. The VT Anechoic Stability Wind Tunnel allowed noise measurements to be carried out using both a 63-elements microphone phased array and a linear array of 15 microphones. The noise sources were identified from the flyover view under various flow speeds and the phased array positioned in both the near and far-field. The directivity pattern of the landing gear was determined using the linear array of microphones. The effectiveness of 4 passive noise control devices was evaluated. The 26%-scale model tested was a faithful reproduction of the full-scale landing gear and included most of the full-scale details with accuracy down to 3 mm. The same landing gear model was previously tested in the original hard-walled configuration of the VT tunnel with the same phased array mounted on the wall of the test section, i.e. near-field position. Thus, the new anechoic configuration of the VT wind tunnel offered a unique opportunity to directly compare, using the same gear model and phased array instrumentation, data collected in hard-walled and semi-anechoic test sections. The main objectives of the present work were (i) to evaluate the validity of conducting aeroacoustic studies in non-acoustically treated, hard-walled wind tunnels, (ii) to test the effectiveness of various streamlining devices (passive noise control) at different flyover locations, and (iii) to assess if phased array measurements can be used to estimate noise reduction. As expected, the results from this work show that a reduction of the background noise (e.g. anechoic configuration) leads to significantly cleaner beamforming maps and allows one to locate noise sources that would not be identified otherwise. By using the integrated spectra for the baseline landing gear, it was found that in the hard-walled test section the levels of the landing gear noise were overestimated. Phased array measurements in the near and far-field positions were also compared in the anechoic configuration. The results showed that straight under the gear, near-field measurements located only the lower-truck noise sources, i.e. noise components located behind the truck were shielded. It was thus demonstrated that near-field, phased-array measurements of the landing gear noise straight under the gear are not suitable. The array was also placed in the far-field, on the rear-arc of the landing gear. From this position, other noise sources such as the strut could be identified. This result demonstrated that noise from the landing gear on the flyover path cannot be characterized by only taking phased array measurement right under the gear. The noise reduction potential of various streamlining devices was estimated from phased array measurements (by integrating the beamforming maps) and using the linear array of individually calibrated microphones. Comparison of the two approaches showed that the reductions estimated from the phased array and a single microphone were in good agreement in the far-field. However, it was found that in the near-field, straight under the gear, phased array measurements greatly overestimate the attenuation. / Master of Science

[EN] The aim of this work is to design new acoustic devices based on arrangements of scattering units. First, the use of sonic crystals as noise barriers for traffic noise control is comprehensively analyzed. Due to the limitations of the conventional structures based on rigid scatterers, the inclusion of absorbing elements is proposed. Two different types of absorbers are here considered: porous materials and microperforated plates. In the first case, the attenuation characteristics of barriers made with cylinders containing rubber crumb is analyzed. The second proposal is based on the construction of cylindrical microperforated shells. Analytical approaches modelling the behavior of the barriers have been developed in both cases. These models show a satisfactory agreement with the corresponding experimental realizations. Finally, it is performed an optimization process in order to obtain efficient sound barriers intended to attenuate traffic noise. Another application considered in this work is the construction of cloaks to render objects acoustically invisible. In particular, cloaks made with rigid inclusions are designed to operate with airborne sound. The first proposal consists of a cloak that utilizes the temperature of the background to control the properties of the effective medium. In addition, two and three-dimensional cloaks have been developed through the scattering cancellation technique. These devices have been designed by means of an optimization procedure and their performance has been experimentally demonstrated. / [ES] El objetivo de este trabajo es el diseño de nuevos dispositivos acústicos basados en disposiciones de centros de dispersión. En primer lugar, el uso de cristales sónicos como barreras acústicas para el control de ruido de tráfico es analizado en detalle. Debido a las limitaciones que presentan las estructuras convencionales basadas en centros de dispersión rígidos, se propone la inclusión de elementos absorbentes en los mismos. Se han considerado dos tipos distintos de absorbente: materiales porosos y placas microperforadas. En el primer caso se analizan las propiedades atenuadoras de barreras formadas por cilindros que contienen granza de caucho. La segunda solución se basa en la construcción de coronas microperforadas. En ambos casos se han desarrollado modelos analíticos que permiten determinar el comportamiento de las barreras. Dichos modelos muestran un acuerdo satisfactorio con las correspondientes realizaciones experimentales. Finalmente se ha realizado un proceso de optimización con objeto de obtener barreras eficientes para la atenuación de ruido de tráfico. Otra aplicación considerada en este trabajo es el desarrollo de dispositivos de invisibilidad acústica. Concretamente se pretenden diseñar mantos constituidos con elementos rígidos para ondas acústicas en aire. La primera propuesta consiste en un manto que utiliza la temperatura del medio externo para controlar sus propiedades efectivas. Además se han desarrollado mantos en dos y tres dimensiones a través de la técnica de cancelación de la dispersión. Los diseños han sido realizados por medio de un proceso de optimización y su funcionamiento ha sido demostrado experimentalmente. / [CA] L'objectiu d'aquest treball és el disseny de nous dispositius acústics basats en disposicions de centres de dispersió. En primer lloc, l'ús de vidres sònics com barreres acústiques per al control de soroll de trànsit és analitzat en detall. A causa de les limitacions que presenten les estructures convencionals basades en centres de dispersió rígids, es proposa la inclusió d'elements absorbents en els mateixos. S'han considerat dos tipus diferents de absorbent: materials porosos i plaques microperforades. En el primer cas s'analitzen les propietats atenuadores de barreres formades per cilindres que contenen gransa de cautxú. La segona solució es basa en la construcció de corones microperforades. En tots dos casos s'han desenvolupat models analítics que permeten determinar el comportament de les barreres. Aquests models mostren un acord satisfactori amb les corresponents realitzacions experimentals. Finalment s'ha realitzat un procés d'optimització per tal d'obtenir barreres eficients per l'atenuació de soroll de trànsit. Una altra aplicació considerada en aquest treball és el desenvolupament de dispositius d'invisibilitat acústica. Concretament es pretenen dissenyar mantells constituïts amb elements rígids per ones acústiques en aire. La primera proposta consisteix en un mantell que utilitza la temperatura del medi extern per controlar les seves propietats efectives. A més s'han desenvolupat mantells en dues i tres dimensions a través de la tècnica de cancel·lació de la dispersió. Els dissenys han estat realitzats per mitjà d'un procés d'optimització i el seu funcionament ha estat demostrat experimentalment. / García Chocano, VM. (2015). New devices for noise control and acoustic cloaking [Tesis doctoral]. Editorial Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/53026 / Premios Extraordinarios de tesis doctorales

[EN] The field of audio signal processing has undergone a major development in recent years. Both the consumer and professional marketplaces continue to show growth in audio applications such as immersive audio schemes that offer optimal listening experience, intelligent noise reduction in cars or improvements in audio teleconferencing or hearing aids. The development of these applications has a common interest in increasing or improving the number of discrete audio channels, the quality of the audio or the sophistication of the algorithms. This often gives rise to problems of high computational cost, even when using common signal processing algorithms, mainly due to the application of these algorithms to multiple signals with real-time requirements. The field of High Performance Computing (HPC) based on low cost hardware elements is the bridge needed between the computing problems and the real multimedia signals and systems that lead to user's applications. In this sense, the present thesis goes a step further in the development of these systems by using the computational power of General Purpose Graphics Processing Units (GPGPUs) to exploit the inherent parallelism of signal processing for multichannel audio applications. The increase of the computational capacity of the processing devices has been historically linked to the number of transistors in a chip. However, nowadays the improvements in the computational capacity are mainly given by increasing the number of processing units and using parallel processing. The Graphics Processing Units (GPUs), which have now thousands of computing cores, are a representative example. The GPUs were traditionally used to graphic or image processing, but new releases in the GPU programming environments such as CUDA have allowed the use of GPUS for general processing applications. Hence, the use of GPUs is being extended to a wide variety of intensive-computation applications among which audio processing is included. However, the data transactions between the CPU and the GPU and viceversa have questioned the viability of the use of GPUs for audio applications in which real-time interaction between microphones and loudspeakers is required. This is the case of the adaptive filtering applications, where an efficient use of parallel computation in not straightforward. For these reasons, up to the beginning of this thesis, very few publications had dealt with the GPU implementation of real-time acoustic applications based on adaptive filtering. Therefore, this thesis aims to demonstrate that GPUs are totally valid tools to carry out audio applications based on adaptive filtering that require high computational resources. To this end, different adaptive applications in the field of audio processing are studied and performed using GPUs. This manuscript also analyzes and solves possible limitations in each GPU-based implementation both from the acoustic point of view as from the computational point of view. / [ES] El campo de procesado de señales de audio ha experimentado un desarrollo importante en los últimos años. Tanto el mercado de consumo como el profesional siguen mostrando un crecimiento en aplicaciones de audio, tales como: los sistemas de audio inmersivo que ofrecen una experiencia de sonido óptima, los sistemas inteligentes de reducción de ruido en coches o las mejoras en sistemas de teleconferencia o en audífonos. El desarrollo de estas aplicaciones tiene un propósito común de aumentar o mejorar el número de canales de audio, la propia calidad del audio o la sofisticación de los algoritmos. Estas mejoras suelen dar lugar a sistemas de alto coste computacional, incluso usando algoritmos comunes de procesado de señal. Esto se debe principalmente a que los algoritmos se suelen aplicar a sistemas multicanales con requerimientos de procesamiento en tiempo real. El campo de la Computación de Alto Rendimiento basado en elementos hardware de bajo coste es el puente necesario entre los problemas de computación y los sistemas multimedia que dan lugar a aplicaciones de usuario. En este sentido, la presente tesis va un paso más allá en el desarrollo de estos sistemas mediante el uso de la potencia de cálculo de las Unidades de Procesamiento Gráfico (GPU) en aplicaciones de propósito general. Con ello, aprovechamos la inherente capacidad de paralelización que poseen las GPU para procesar señales de audio y obtener aplicaciones de audio multicanal. El aumento de la capacidad computacional de los dispositivos de procesado ha estado vinculado históricamente al número de transistores que había en un chip. Sin embargo, hoy en día, las mejoras en la capacidad computacional se dan principalmente por el aumento del número de unidades de procesado y su uso para el procesado en paralelo. Las GPUs son un ejemplo muy representativo. Hoy en día, las GPUs poseen hasta miles de núcleos de computación. Tradicionalmente, las GPUs se han utilizado para el procesado de gráficos o imágenes. Sin embargo, la aparición de entornos sencillos de programación GPU, como por ejemplo CUDA, han permitido el uso de las GPU para aplicaciones de procesado general. De ese modo, el uso de las GPU se ha extendido a una amplia variedad de aplicaciones que requieren cálculo intensivo. Entre esta gama de aplicaciones, se incluye el procesado de señales de audio. No obstante, las transferencias de datos entre la CPU y la GPU y viceversa pusieron en duda la viabilidad de las GPUs para aplicaciones de audio en las que se requiere una interacción en tiempo real entre micrófonos y altavoces. Este es el caso de las aplicaciones basadas en filtrado adaptativo, donde el uso eficiente de la computación en paralelo no es sencillo. Por estas razones, hasta el comienzo de esta tesis, había muy pocas publicaciones que utilizaran la GPU para implementaciones en tiempo real de aplicaciones acústicas basadas en filtrado adaptativo. A pesar de todo, esta tesis pretende demostrar que las GPU son herramientas totalmente válidas para llevar a cabo aplicaciones de audio basadas en filtrado adaptativo que requieran elevados recursos computacionales. Con este fin, la presente tesis ha estudiado y desarrollado varias aplicaciones adaptativas de procesado de audio utilizando una GPU como procesador. Además, también analiza y resuelve las posibles limitaciones de cada aplicación tanto desde el punto de vista acústico como desde el punto de vista computacional. / [CA] El camp del processament de senyals d'àudio ha experimentat un desenvolupament important als últims anys. Tant el mercat de consum com el professional segueixen mostrant un creixement en aplicacions d'àudio, com ara: els sistemes d'àudio immersiu que ofereixen una experiència de so òptima, els sistemes intel·ligents de reducció de soroll en els cotxes o les millores en sistemes de teleconferència o en audiòfons. El desenvolupament d'aquestes aplicacions té un propòsit comú d'augmentar o millorar el nombre de canals d'àudio, la pròpia qualitat de l'àudio o la sofisticació dels algorismes que s'utilitzen. Això, sovint dóna lloc a sistemes d'alt cost computacional, fins i tot quan es fan servir algorismes comuns de processat de senyal. Això es deu principalment al fet que els algorismes se solen aplicar a sistemes multicanals amb requeriments de processat en temps real. El camp de la Computació d'Alt Rendiment basat en elements hardware de baix cost és el pont necessari entre els problemes de computació i els sistemes multimèdia que donen lloc a aplicacions d'usuari. En aquest sentit, aquesta tesi va un pas més enllà en el desenvolupament d'aquests sistemes mitjançant l'ús de la potència de càlcul de les Unitats de Processament Gràfic (GPU) en aplicacions de propòsit general. Amb això, s'aprofita la inherent capacitat de paral·lelització que posseeixen les GPUs per processar senyals d'àudio i obtenir aplicacions d'àudio multicanal. L'augment de la capacitat computacional dels dispositius de processat ha estat històricament vinculada al nombre de transistors que hi havia en un xip. No obstant, avui en dia, les millores en la capacitat computacional es donen principalment per l'augment del nombre d'unitats de processat i el seu ús per al processament en paral·lel. Un exemple molt representatiu són les GPU, que avui en dia posseeixen milers de nuclis de computació. Tradicionalment, les GPUs s'han utilitzat per al processat de gràfics o imatges. No obstant, l'aparició d'entorns senzills de programació de la GPU com és CUDA, han permès l'ús de les GPUs per a aplicacions de processat general. D'aquesta manera, l'ús de les GPUs s'ha estès a una àmplia varietat d'aplicacions que requereixen càlcul intensiu. Entre aquesta gamma d'aplicacions, s'inclou el processat de senyals d'àudio. No obstant, les transferències de dades entre la CPU i la GPU i viceversa van posar en dubte la viabilitat de les GPUs per a aplicacions d'àudio en què es requereix la interacció en temps real de micròfons i altaveus. Aquest és el cas de les aplicacions basades en filtrat adaptatiu, on l'ús eficient de la computació en paral·lel no és senzilla. Per aquestes raons, fins al començament d'aquesta tesi, hi havia molt poques publicacions que utilitzessin la GPU per implementar en temps real aplicacions acústiques basades en filtrat adaptatiu. Malgrat tot, aquesta tesi pretén demostrar que les GPU són eines totalment vàlides per dur a terme aplicacions d'àudio basades en filtrat adaptatiu que requereixen alts recursos computacionals. Amb aquesta finalitat, en la present tesi s'han estudiat i desenvolupat diverses aplicacions adaptatives de processament d'àudio utilitzant una GPU com a processador. A més, aquest manuscrit també analitza i resol les possibles limitacions de cada aplicació, tant des del punt de vista acústic, com des del punt de vista computacional. / Lorente Giner, J. (2015). Adaptive signal processing for multichannel sound using high performance computing [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/58427