Master´s work discusses the issue of processing audio signal by means of signal processor. At present, the signal processor occurs in almost all devices that process or somehow modify sound in digital form. The aim of this master´s thesis is to study the signal processors from different producers, which are now on the market and select one suitable type for device that will process the audio signal. With this signal processor then propose a circuit diagram for a device that will process the audio signal and the PSpice simulation of the circuit and construct this device.

The aim of this work was to design and construct a simple audio amplifier, which would be based of involvement and components used in switching power sources on the maximum degree (for example from usual ATX sources for computer). The main motivation for development is the existence of many old sources that do not use and big interest of users for the construction of home audio amplifiers for small and medium power. Amplifiers based on switching power sources with greater efficiency and less weight than comparable classic amplifiers (working in classes A, B or AB). Their disadvantage is the complicated design and introduction of distortion and spurious signals to the amplified signal. This work will be described a design of simple amplifier, which will minimize these negative effects. For shortcomings and wrong parameters this proposed involvement will be eventually described and implemented design of amplifier in class D.

The main content of this masters’s thesis is designing of an audio mixing desk and simulation of individual components in software OrCAD. The most important parts of the device are input preamplifiers for dynamic, electret and condenser microphones, stereo unbalanced inputs and balanced line level preamplifiers, equalization circuits and LED level indicators of individual channels, headphone listening circuit, 10-band equalizer, audio spectrum analyzer, circuits with balanced signal for main outputs and power supply circuits.

This paper deals with development and construction of an electronic drum module. The purpose of the device is to capture signals from an electronic drumkit and to produce sound accordingly. Firstly, a protoype with no sound output is constructed to demonstrate an ability to capture input signals. Based on its function, parameters for a final device are determined. Then, electronic component selection and design of printed circuit boards is described with an aim to maximalize polyphony and minimize latency of the sound generator. After that, firmware with software mixing algorithm is designed. Lastly, testing and measurement of real device parameters is performed.

V oblasti zpracování signálů se v současné době čím dál více využívají tzv. řídké reprezentace signálů, tzn. že daný signál je možné vyjádřit přesně či velmi dobře aproximovat lineární kombinací velmi malého počtu vektorů ze zvoleného reprezentačního systému. Tato práce se zabývá využitím řídkých reprezentací pro rekonstrukci poškozených zvukových záznamů, ať už historických nebo nově vzniklých. Především historické zvukové nahrávky trpí zarušením jako praskání nebo šum. Krátkodobé poškození zvukových nahrávek bylo doposud řešeno interpolačními technikami, zejména pomocí autoregresního modelování. V nedávné době byl představen algoritmus s názvem Audio Inpainting, který řeší doplňování chybějících vzorků ve zvukovém signálu pomocí řídkých reprezentací. Zmíněný algoritmus využívá tzv. hladové algoritmy pro řešení optimalizačních úloh. Cílem této práce je porovnání dosavadních interpolačních metod s technikou Audio Inpaintingu. Navíc, k řešení optimalizačních úloh jsou využívány algoritmy založené na l1-relaxaci, a to jak ve formě analyzujícího, tak i syntetizujícího modelu. Především se jedná o proximální algoritmy. Tyto algoritmy pracují jak s jednotlivými koeficienty samostatně, tak s koeficienty v závislosti na jejich okolí, tzv. strukturovaná řídkost. Strukturovaná řídkost je dále využita taky pro odšumování zvukových nahrávek. Jednotlivé algoritmy jsou v praktické části zhodnoceny z hlediska nastavení parametrů pro optimální poměr rekonstrukce vs. výpočetní čas. Všechny algoritmy popsané v práci jsou na praktických příkladech porovnány pomocí objektivních metod odstupu signálu od šumu (SNR) a PEMO-Q. Na závěr je úspěšnost rekonstrukce poškozených zvukových signálů vyhodnocena.

Bien que les réseaux de neurones soient à présent utilisés dans la quasi-totalité des composants d’un système de reconnaissance de la parole, du modèle acoustique au modèle de langue, l’entrée de ces systèmes reste une représentation analytique et fixée de la parole dans le domaine temps-fréquence, telle que les mel-filterbanks. Cela se distingue de la vision par ordinateur, un domaine où les réseaux de neurones prennent en entrée les pixels bruts. Les mel-filterbanks sont le produit d’une connaissance précieuse et documentée du système auditif humain, ainsi que du traitement du signal, et sont utilisées dans les systèmes de reconnaissance de la parole les plus en pointe, systèmes qui rivalisent désormais avec les humains dans certaines conditions. Cependant, les mel-filterbanks, comme toute représentation fixée, sont fondamentalement limitées par le fait qu’elles ne soient pas affinées par apprentissage pour la tâche considérée. Nous formulons l’hypothèse qu’apprendre ces représentations de bas niveau de la parole, conjontement avec le modèle, permettrait de faire avancer davantage l’état de l’art. Nous explorons tout d’abord des approches d’apprentissage faiblement supervisé et montrons que nous pouvons entraîner un unique réseau de neurones à séparer l’information phonétique de celle du locuteur à partir de descripteurs spectraux ou du signal brut et que ces représentations se transfèrent à travers les langues. De plus, apprendre à partir du signal brut produit des représentations du locuteur significativement meilleures que celles d’un modèle entraîné sur des mel-filterbanks. Ces résultats encourageants nous mènent par la suite à développer une alternative aux mel-filterbanks qui peut être entraînée à partir des données. Dans la seconde partie de cette thèse, nous proposons les Time-Domain filterbanks, une architecture neuronale légère prenant en entrée la forme d’onde, dont on peut initialiser les poids pour répliquer les mel-filterbanks et qui peut, par la suite, être entraînée par rétro-propagation avec le reste du réseau de neurones. Au cours d’expériences systématiques et approfondies, nous montrons que les Time-Domain filterbanks surclassent systématiquement les melfilterbanks, et peuvent être intégrées dans le premier système de reconnaissance de la parole purement convolutif et entraîné à partir du signal brut, qui constitue actuellement un nouvel état de l’art. Les descripteurs fixes étant également utilisés pour des tâches de classification non-linguistique, pour lesquelles elles sont d’autant moins optimales, nous entraînons un système de détection de dysarthrie à partir du signal brut, qui surclasse significativement un système équivalent entraîné sur des mel-filterbanks ou sur des descripteurs de bas niveau. Enfin, nous concluons cette thèse en expliquant en quoi nos contributions s’inscrivent dans une transition plus large vers des systèmes de compréhension du son qui pourront être appris de bout en bout. / While deep neural networks are now used in almost every component of a speech recognition system, from acoustic to language modeling, the input to such systems are still fixed, handcrafted, spectral features such as mel-filterbanks. This contrasts with computer vision, in which a deep neural network is now trained on raw pixels. Mel-filterbanks contain valuable and documented prior knowledge from human auditory perception as well as signal processing, and are the input to state-of-the-art speech recognition systems that are now on par with human performance in certain conditions. However, mel-filterbanks, as any fixed representation, are inherently limited by the fact that they are not fine-tuned for the task at hand. We hypothesize that learning the low-level representation of speech with the rest of the model, rather than using fixed features, could push the state-of-the art even further. We first explore a weakly-supervised setting and show that a single neural network can learn to separate phonetic information and speaker identity from mel-filterbanks or the raw waveform, and that these representations are robust across languages. Moreover, learning from the raw waveform provides significantly better speaker embeddings than learning from mel-filterbanks. These encouraging results lead us to develop a learnable alternative to mel-filterbanks, that can be directly used in replacement of these features. In the second part of this thesis we introduce Time-Domain filterbanks, a lightweight neural network that takes the waveform as input, can be initialized as an approximation of mel-filterbanks, and then learned with the rest of the neural architecture. Across extensive and systematic experiments, we show that Time-Domain filterbanks consistently outperform melfilterbanks and can be integrated into a new state-of-the-art speech recognition system, trained directly from the raw audio signal. Fixed speech features being also used for non-linguistic classification tasks for which they are even less optimal, we perform dysarthria detection from the waveform with Time-Domain filterbanks and show that it significantly improves over mel-filterbanks or low-level descriptors. Finally, we discuss how our contributions fall within a broader shift towards fully learnable audio understanding systems.

This diploma thesis are describing the evolution of audio commentary in sports computer games since 1998. As the subject for the research was selected the FIFA game series, currently the best-selling game of virtual sports category. The objective of the analysis is to find out under which circumstances could this game includes the Czech version of audio commentary. Structurally this diploma thesis is divided into theoretical and research part. The first part describes the theoretical aspects of the researched issue. Describes the evolution in the video game industry from its beginnings to the present time, including the phenomenon of the eSports that is described in the second chapter. The theoretical part concludes with an analysis of methods, elements and specifics, for commentating of a sporting events. At the beginning of the research, this analysis is confronted with the specifics of the word algorithm for audio commentary used in the FIFA game series. Final SWOT analysis describe evaluation of the current state of evolution in the video game industry with the focus on the subject of this diploma thesis, audio commentary in the FIFA game series and possible return of the Czech version of audio commentary into the game. Keywords Commentating, FIFA, pc games, videogames, word algortihm,...

"Most stations probably keep various materials in order that they may be used again. Because there is a definite saving of time if materials are stored systematically and are readily available, the question arises as to how the various educational television stations in the country maintain a record of what they have produced and obtained in the way of audio-visual materials which will probably be useful in later programming. The purpose of this study, therefore, is to: (1) determine what types of audio-visual materials educational television stations store; (2) learn of the physical arrangement of the storage of these materials; and (3) ascertain if these materials are indexed, and, if so, in what manner. In order to answer these questions, it was decided that a survey of the methods of storage of audio-visual materials in educational stations at least five years old would be attempted"--Introduction. / Typescript. / "June, 1961." / "Submitted to the Graduate School of Florida State University in partial fulfillment of the requirements for the degree of Master of Science." / Advisor: Ruth H. Rockwood, Professor Directing Paper. / Includes bibliographical references (leaves 42-43).

Målet med det här arbetet har varit att utveckla en ljudmotor skriven i JavaScript som underlättar ljuddesign för webbsidor. Ett sekundärt mål har varit att diskutera hur relevant Flash är för ljuddesign på webben och huruvida HTML5 har möjligheten att ersätta Flash i framtiden. En ljudmotor som har fått namnet Klang har utvecklats som ett JavaScriptbibliotek för integration med interaktiva webbsidor. Ljudmotorn använder Web Audio API, den nyligen föreslådda ljudstandarden för HTML5, som bas för uppspelning och processering av ljud. För att anpassa ljuddesignen för olika typer av webbsidor har Klang möjligheten att dynamiskt styra hur musik och ljud spelas upp beroende på vilken typ av webbsida den används för och hur en besökare använder denna webbsida. Klang kommer att behöva fortsatt underhåll i framtiden när Web Audio API växer men ger en god överblick av vad som är ljudmässigt möjligt med HTML5 utan externa plugin. Rapportens slutsats förespråkar utvecklingen av HTML5 och möjligheterna det har infört ur ett ljudperspektiv men medger att Flash än så länge behåller sin plats i webbutvecklarens repertoar. / The primary objective of this work has been developing a JavaScript audio engine to aid in designing audio for the web. This project also covers discussion of whether or not HTML5 may replace Flash for the purpose of sound design for the web in the future. Klang, an audio engine available as a JavaScript plugin has been developed to be integrated with interactive websites. The audio engine uses Web Audio API, the newly proposed audio standard for HTML5, for audio playback and processing. To be able to adapt the audio engine for the web site it is used for, Klang offers a way to dynamically control the behaviour of music and sounds according to the type of web site it is used for and how users interact with the web site. Klang will be in need of updates as the Web Audio API continues to develop but demonstrates the potential of the audio features of HTML5. The conclusion of the report advocates the use of pure HTML5 and the improvements it brings to the web's audio department but recognizes that Flash still has a place in the repertoire of the web developer.

Music in general and recorded music in particular are rarely a priority for libraries’ digitization policies, although wax cylinders and 78rpm discs might be digitized for preservation and accessibility reasons. The respect of the original recording technique during the digitization process will ensure the scientific and artistic credibility of the digitized sources. The Royal Library of Belgium started in 2016 the digitization of its whole collection of 78rpm. Realized by subcontracting, this project of about 4,000 hours will constitute a large corpus of sources for the digital musicology upcoming needs. / Musikalien im Allgemeinen und Tonträger im Besonderen erhalten in bibliothekarischen Digitalisierungsstrategien häufig nur wenig Beachtung, obwohl gerade für Wachszylinder und Schellackplatten sowohl aus Gründen der Bestandserhaltung als auch zur Verbesserung der Zugänglichkeit ihre Digitalisierung dringend geboten wäre. Um bei der Retrodigitalisierung von historischen Tonaufnahmen künstlerisch und wissenschaftlich zuverlässige Ergebnisse zu erreichen, ist den originalen Aufnahmetechniken große Aufmerksamkeit zu schenken. Die Königliche Bibliothek Belgiens lässt seit 2016 ihre vollständige Sammlung an Schellackplatten digitalisieren. Mithilfe eines Dienstleisters werden ca. 4 000 Stunden Tonaufnahmen produziert, die für aufkommende Forschungsfragen der digitalen Musikwissenschaft einen gewichtigen Quellenkorpus darstellen.