Dissertationen zum Thema „Audio content analysis“

The information age has brought with it a dual problem. In the first place, the ready access to mechanisms to capture and store vast amounts of data in all forms (text, audio, image and video), has resulted in a continued demand for ever more efficient means to store and transmit this data. In the second, the rapidly increasing store demands effective means to structure and access the data in an efficient and meaningful manner. In terms of audio data, the first challenge has traditionally been the realm of audio compression research that has focused on statistical, unstructured audio representations that obfuscate the inherent structure and semantic content of the underlying data. This has only served to further complicate the resolution of the second challenge resulting in access mechanisms that are either impractical to implement, too inflexible for general application or too low level for the average user. Thus, an artificial dichotomy has been created from what is in essence a dual problem. The founding motivation of this thesis is that, although the hypermedia model has been identified as the ideal, cognitively justified method for organising data, existing audio data representations and coding models provide little, if any, support for, or resemblance to, this model. It is the contention of the author that any successful attempt to create hyperaudio must resolve this schism, addressing both storage and information management issues simultaneously. In order to achieve this aim, an audio representation must be designed that provides compact data storage while, at the same time, revealing the inherent structure of the underlying data. Thus it is the aim of this thesis to present a representation designed with these factors in mind. Perhaps the most difficult hurdle in the way of achieving the aims of content-based audio coding and information management is that of auditory source separation. The MPEG committee has noted this requirement during the development of its MPEG-7 standard, however, the mechanics of "how" to achieve auditory source separation were left as an open research question. This same committee proposed that MPEG-7 would "support descriptors that can act as handles referring directly to the data, to allow manipulation of the multimedia material." While meta-data tags are a part solution to this problem, these cannot allow manipulation of audio material down to the level of individual sources when several simultaneous sources exist in a recording. In order to achieve this aim, the data themselves must be encoded in such a manner that allows these descriptors to be formed. Thus, content-based coding is obviously required. In the case of audio, this is impossible to achieve without effecting auditory source separation. Auditory source separation is the concern of computational auditory scene analysis (CASA). However, the findings of CASA research have traditionally been restricted to a limited domain. To date, the only real application of CASA research to what could loosely be classified as information management has been in the area of signal enhancement for automatic speech recognition systems. In these systems, a CASA front end serves as a means of separating the target speech from the background "noise". As such, the design of a CASA-based approach, as presented in this thesis, to one of the most significant challenges facing audio information management research represents a significant contribution to the field of information management. Thus, this thesis unifies research from three distinct fields in an attempt to resolve some specific and general challenges faced by all three. It describes an audio representation that is based on a sinusoidal model from which low-level auditory primitive elements are extracted. The use of a sinusoidal representation is somewhat contentious with the modern trend in CASA research tending toward more complex approaches in order to resolve issues relating to co-incident partials. However, the choice of a sinusoidal representation has been validated by the demonstration of a method to resolve many of these issues. The majority of the thesis contributes several algorithms to organise the low-level primitives into low-level auditory objects that may form the basis of nodes or link anchor points in a hyperaudio structure. Finally, preliminary investigations in the representation’s suitability for coding and information management tasks are outlined as directions for future research.

Women have historically been the subject of stereotypes – especially criminalized women as they are constructed in the mass media. These stereotypes become particularly problematic when they are invoked in infotainment media – a genre that combines information and entertainment and presents itself as primarily factual. As such, ideological messages delivered through infotainment are also (re)presented as truthful and may be more likely to be taken up by an unquestioning audience. This research aimed to answer the following research question: How does infotainment portray women who commit serious violent crime? In order to answer this question, a qualitative content analysis was employed and “Deadly Women”, a televised infotainment series that narrates and re-enacts true crime stories of women who kill, was selected as a case study. The sample consisted of previously identified typologies: mothers who kill their children, women who kill their partners, adolescent girls who kill, and vigilantes who kill their abusers. Stemming from a critical feminist framework, the analysis revealed that Deadly Women relies on two primary trajectories to explain the violence committed by women and girls. While both trajectories emphasized gendered stereotypes that involved emotionality and mental health issues, they were nonetheless distinct. The first trajectory evoked narratives of the ‘emotionless’ and ‘psychopathic’ perpetrator; while the second trajectory characterized the offender as overly ‘emotional’ and ‘depressed’. These trajectories, along with their related variables, problematically (re)presented violent women and girls in simplistic and dualistic manners that served to obscure rather than to clarify the circumstances surrounding their crimes.

Aquesta tesi doctoral proposa i avalua un enfocament computacional per a la descripció automàtica dels aspectes tonals de la música a partir de l'anàlisi de senyals d'-audio polifòniques. Aquests mètodes es centren en el càlcul de descriptors de distribucions de notes, en l'estimació de tonalitat d'una peça, en la visualització de l'evolució del centre tonal o en la mesura de la similitud tonal entre dues peces diferents.
Aquesta tesi contribueix substancialment al camp de la descripció tonal mitjançant mètodes computacionals: a) Proporciona una revisió multidisciplinària dels sistemes d'estimació de la tonalitat; b) Defineix una sèrie de requeriments que han de complir els descriptors tonals de baix nivell; c) Proporciona una avaluació quantitativa i modular dels mètodes proposats; d) Justifica la idea de que per a certes aplicacions es poden fer servir mètodes que treballen amb partitures sense la necessitat de realitzar una transcripció automàtica e) Estén la literatura existent que treballa amb música clàssica a altres generes musicals; f) Demostra la utilitat dels descriptors tonals per a comparar peces musicals; g) Proporciona un algoritme optimitzat que es fa servir dins un sistema real per a visualització, cerca i recomanació musical, que treballa amb més d'un milió de obres musicals.
Esta tesis doctoral propone y evalúa un enfoque computacional para la descripción automática de aspectos tonales de la música a partir del análisis de señales de audio polifónicas. Estos métodos se centran en calcular descriptores de distribución de notas, en estimar la tonalidad de una pieza, en visualizar la evolución del centro tonal o en medir la similitud tonal entre dos piezas diferentes.
Esta tesis contribuye sustancialmente al campo de la descripción tonal mediante métodos computacionales: a) Proporciona una revisión multidisciplinar de los sistemas de estimación de la tonalidad; b) Define una serie de requerimientos que deben cumplir los descriptores tonales de bajo nivel; c) Proporciona una evaluación cuantitativa y modular de los métodos propuestos; d) Respalda la idea de que para ciertas aplicaciones no es necesario obtener una transcripción perfecta de la partitura, y que se pueden utilizar métodos que trabajan con partituras sin realizar una transcripción automática; e) Extiende la literatura existente que trabaja con música clásica a otros géneros musicales; f) Demuestra la utilidad de los descriptores tonales para comparar piezas musicales; g) Proporciona un algoritmo optimizado que se utiliza en un sistema real para visualización, búsqueda y recomendación musical, que trabaja con mas de un millón de piezas musicales.
This doctoral dissertation proposes and evaluates a computational approach for the automatic description of tonal aspects of music from the analysis of polyphonic audio signals. These algorithms focus on the computation of pitch class distributions descriptors, the estimation of the key of a piece, the visualization of the evolution of its tonal center or the measurement of the similarity between two different musical pieces.
This dissertation substantially contributes to the field of computational tonal description: a) It provides a multidisciplinary review of tonal induction systems; b) It defines a set of requirements for low-level tonal features; c) It provides a quantitative and modular evaluation of the proposed methods; d) It contributes to bridge the gap between audio and symbolic-oriented methods without the need of a perfect transcription; e) It extents current literature dealing with classical music to other musical genres; f) It shows the usefulness of tonal descriptors for music similarity; g) It provides an optimized method which is used in a real system for music visualization and retrieval, working with over a million of musical pieces.

This thesis examines the possible benefits of adapting a biologically-inspired model of human auditory processing as part of a machine-hearing system. Features were generated by an auditory model, and used as input to machine learning systems to determine the content of the sound. Features were generated using the auditory image model (AIM) and were used for speech recognition and audio search. AIM comprises processing to simulate the human cochlea, and a 'strobed temporal integration' process which generates a stabilised auditory image (SAI) from the input sound. The communication sounds which are produced by humans, other animals, and many musical instruments take the form of a pulse-resonance signal: pulses excite resonances in the body, and the resonance following each pulse contains information both about the type of object producing the sound and its size. In the case of humans, vocal tract length (VTL) determines the size properties of the resonance. In the speech recognition experiments, an auditory filterbank was combined with a Gaussian fitting procedure to produce features which are invariant to changes in speaker VTL. These features were compared against standard mel-frequency cepstral coefficients (MFCCs) in a size-invariant syllable recognition task. The VTL-invariant representation was found to produce better results than MFCCs when the system was trained on syllables from simulated talkers of one range of VTLs and tested on those from simulated talkers with a different range of VTLs. The image stabilisation process of strobed temporal integration was analysed. Based on the properties of the auditory filterbank being used, theoretical constraints were placed on the properties of the dynamic thresholding function used to perform strobe detection. These constraints were used to specify a simple, yet robust, strobe detection algorithm. The syllable recognition system described above was then extended to produce features from profiles of the SAI and tested with the same syllable database as before. For clean speech, performance of the features was comparable to that of those generated from the filterbank output. However when pink noise was added to the stimuli, performance dropped more slowly as a function of signal-to-noise ratio when using the SAI-based AIM features, than when using either the filterbank-based features or the MFCCs, demonstrating the noise-robustness properties of the SAI representation. The properties of the auditory filterbank in AIM were also analysed. Three models of the cochlea were considered: the static gammatone filterbank, dynamic compressive gammachirp (dcGC) and the pole-zero filter cascade (PZFC). The dcGC and gammatone are standard filterbank models, whereas the PZFC is a filter cascade, which more accurately models signal propagation in the cochlea. However, while the architecture of the filterbanks is different, they have all been successfully fitted to psychophysical masking data from humans. The abilities of the filterbanks to measure pitch strength were assessed, using stimuli which evoke a weak pitch percept in humans, in order to ascertain whether there is any benefit in the use of the more computationally efficient PZFC.Finally, a complete sound effects search system using auditory features was constructed in collaboration with Google research. Features were computed from the SAI by sampling the SAI space with boxes of different scales. Vector quantization (VQ) was used to convert this multi-scale representation to a sparse code. The 'passive-aggressive model for image retrieval' (PAMIR) was used to learn the relationships between dictionary words and these auditory codewords. These auditory sparse codes were compared against sparse codes generated from MFCCs, and the best performance was found when using the auditory features.

Con la reciente explosión cuantitativa de bibliotecas y colecciones de música en formato
digital, la descripción del contenido desempeña un papel fundamental para una gestión y
búsqueda eficientes de archivos de audio. La presente tesis doctoral pretende hacer un
análisis automático de la estructura de piezas musicales a partir del análisis de una
grabación, es decir, extraer una descripción estructural a partir de señales musicales
polifónicas. En la medida en que la repetición y transformación de la estructura de la
música genera una identificación única de una obra musical, extraer automáticamente
esta información puede vincular entre sí descripciones de bajo y alto nivel de una señal
musical y puede proporcionar al usuario una manera más efectiva de interactuar con un
contenido de audio. Para algunas aplicaciones basadas en contenido, encontrar los límites
de determinados segmentos de una grabación resulta indispensable. Así pues, también se
investiga la segmentación temporal de audio a nivel semántico, al igual que la
identificación de extractos representativos de una señal musical que pueda servir como
resumen de la misma. Para ello se emplea una técnica de análisis a un nivel de
abstracción más elevado que permite obtener una mejor división en segmentos. Tanto
desde el punto de vista teórico como práctico, esta investigación no sólo ayuda a
incrementar nuestro conocimiento respecto a la estructura musical, sino que también
proporciona una ayuda al examen y a la valoración musical.
With the recent explosion in the quantity of digital audio libraries and databases, content
descriptions play an important role in efficiently managing and retrieving audio files.
This doctoral research aims to discover and extract structural description from
polyphonic music signals. As repetition and transformations of music structure creates a
unique identity of music itself, extracting such information can link low-level and higherlevel
descriptions of music signal and provide better quality access plus powerful way of
interacting with audio content. Finding appropriate boundary truncations is indispensable
in certain content-based applications. Thus, temporal audio segmentation at the semantic
level and the identification of representative excerpts from music audio signal are also
investigated. We make use of higher-level analysis technique for better segment
truncation. From both theoretical and practical points of view, this research not only
helps in increasing our knowledge of music structure but also facilitates in time-saving
browsing and assessing of music.

This research investigates how the production and distribution of nationally produced television programmes, films and other audiovisual content can be promoted by regulation. The study identifies current regulatory tools to promote national content and differences in policy approaches. It also identifies new issues resulting from the substantial transformation the media environment has undergone in recent years. The audiovisual sector today is characterised by an abundance of television channels and telecommunication services and by ongoing digital convergence, all impacting on the effectiveness and rationale of content regulation. Focused on the UK, France and Brazil, this comparative research investigates the political, regulatory, socio-cultural, economic, technological and market changes of the communication services in the recent decades and how this has impacted on the provision of national audiovisual content. The starting point of the analysis are the 1980s, when broadcasting started to be gradually liberalised, and it concludes in the 2000s, when information technologies, telecommunications and broadcasting converge into interconnected, complementary and supplementary services. This research offers a three step method of analysis which contributes to a new understanding of the mechanisms and implications involved in the production and distribution of nationally produced content in digital times. In the first step, the various socio-cultural aims of communication services as enacted in legislation are analysed. The second step explores the consequent regulatory tools for the fulfilment of those objectives. The third step provides a market evaluation of the audiovisual industries, which broadcasters and other distribution platforms provide which kind of content, and also looks at audiences‟ viewing preferences to get a better understanding of what type of content should be supported through legislation. As outcomes, this research proposes to academics, policymakers and regulators a new definition for national content and a series of regulatory actions for fostering national audiovisual industries in times of digital convergence.

Different music performances of the same piece may significantly differ from each other. Not only the composer and the score defines the listener’s music experience, but the music performance itself is an integral part of this experience. Four parameter classes can be used to describe a performance objectively: tempo and timing, loudness (dynamics), timbre, and pitch. Each of the individual parameters or their combination can generate a unique characteristic performance. The extraction of such objective parameters is one of the difficulties in the field of Music Performance Analysis and Music Information Retrieval. The submitted work summarizes knowledge and methods from both of the fields. The system is applied to extract data from 31 string quartet performances of 2. movement Lento of String Quartet no. 12 F major (1893) by czech romantic composer Antonín Dvořák (1841–1904).

Dans ce travail, nous avons mis au point une méthode de segmentation des contenus audiovisuels applicable aux appareils de stockage domestiques pour cela nous avons expérimenté un système distribué pour l'analyse du contenu composé de modules individuels d'analyse : les service unit. L'un entre eux a été dédié à la caractérisation des éléments hors contenu, i. E. Les publicités, et offre de bonnes perfermances. Parallélement, nous avons testé différents détecteurs de changement de plans afin de retenir le meilleur d'ente eux pour la suite. Puis, nous avons proposé une étude des règles de production des films, i. E. Grammaire de films, qui a permis de définir les séquences de parallel shot. Nous avons, ainsi, testé quatre méthodes de regroupement basées similarité afin de retenir la meilleure d'entre elles pour la suite. Finalement, nous avons recherché différentes méthodes de détection des frontières de scènes et avons obtenu les meilleurs résultats en combinant une méthode basée couleur avec un critère de longueur de plan. Ce dernier offre des performances justifiant son intégration dans les appareils de stockage grand public.

The aim of this research is to examine the determinants of the level of corporate risk disclosure (CRD) in the interim reports of Swedish non-financial companies. A quantitative research approach is used, the sample data of which consist of 166 firms with 4,849 interim reports over a 10-year period. By utilizing the notion of risk and its definition, I have distinguished three categories of risk, namely risk as uncertainty, risk as threat and risk as opportunity. A systematic content analysis is conducted with the use of a software program, which is specifically designed for this purpose. The number of sentences that contain keywords related to the three risk categories is counted as the total CRD score, which is transformed to the disclosure index. I have examined the impact of firms’ characteristics and corporate governance mechanisms on the level of CRD based on agency theory. The ordinary least squares regression method with  control for fixed year effects is used to analyse the data, which show that firm size and audit committee have a positive relationship with the level of corporate risk disclosure. The result demonstrates also that there is a negative relationship between family ownership and the level of CRD, and an insignificant relationship between leverage and the level of CRD.

This tesis reports on research carried out and published during the last twenty years on different problems of Music Information Retrieval (MIR). We organize the text as a personal account and critical reflection along four hypothesized ages that have shaped the evolution of MIR. In the age of feature extractors, we present work on features to describe sounds and music, especially timbre and tonal aspects. In the age of semantic descriptors work on describing music with high-level concepts, such as mood, instruments, similarities, cover versions or genres, usually inferred with machine learning from annotated collections is reported. In the age of context-aware systems we report on user models for recommendation and for avatar generation, in addition to factors that influence music listening decisions. We finally discuss the possibility of a more recent age of creative systems where MIR features, classifiers, models and evaluation methodologies aid to enhance or expand music creation.
Aquesta tesi informa sobre recerca realitzada i publicada durant els últims vint anys en diferents problemes de Recuperació d'Informació Musical (MIR). Organitzem el text com a visió personal i reflexió crítica i utilitzant quatre hipotètiques edats que han configurat l'evolució del MIR. A l'edat dels extractors de característiques, presentem treballs sobre trets per a descriure sons i música, especialment timbre i aspectes tonals. A l'edat dels descriptors semàntics es treballa en la descripció de música amb conceptes d'alt nivell, com l'estat d'ànim, els instruments, les similituds, les versions musicals o els gèneres, generalment deduïts amb l'aprenentatge automàtic a partir de col·leccions anotades. En l'era dels sistemes sensibles al context, informem sobre models d'usuaris amb l’objectiu de fer recomanacions musicals i generació d'avatars, a més de factors que influeixen en les decisions d'escoltar música. S’esmenta, finalmente, una posible i més recent edat dels sistemes creatius on els descriptors, classificadors, models i metodologies d'avaluació de MIR ajuden a potenciar o ampliar la creació musical.

rezensiertes Werk: Jeffrey Burns: The music of psalms, Proverbs and Job in the Hebrew bible : A revised theory of musical accents in the Hebrew bible ; general analysis, bibliography, table of contents CD with complete text and audio files of musically reconstructed psalms, proverbs and Job, sung by computer speech synthesis. Wiesbaden : Harrassowitz Verlag, 2011. 169 S. ISBN 978-3-447-06191-9

Music was the first mass-market industry to be completely restructured by digital technology, and today we can have access to thousands of tracks stored locally on our smartphone and millions of tracks through cloud-based music services. Given the vast quantity of music at our fingertips, we now require novel ways of describing, indexing, searching and interacting with musical content. In this thesis we focus on a technology that opens the door to a wide range of such applications: automatically estimating the pitch sequence of the melody directly from the audio signal of a polyphonic music recording, also referred to as melody extraction. Whilst identifying the pitch of the melody is something human listeners can do quite well, doing this automatically is highly challenging. We present a novel method for melody extraction based on the tracking and characterisation of the pitch contours that form the melodic line of a piece. We show how different contour characteristics can be exploited in combination with auditory streaming cues to identify the melody out of all the pitch content in a music recording using both heuristic and model-based approaches. The performance of our method is assessed in an international evaluation campaign where it is shown to obtain state-of-the-art results. In fact, it achieves the highest mean overall accuracy obtained by any algorithm that has participated in the campaign to date. We demonstrate the applicability of our method both for research and end-user applications by developing systems that exploit the extracted melody pitch sequence for similarity-based music retrieval (version identification and query-by-humming), genre classification, automatic transcription and computational music analysis. The thesis also provides a comprehensive comparative analysis and review of the current state-of-the-art in melody extraction and a first of its kind analysis of melody extraction evaluation methodology.
La industria de la música fue una de las primeras en verse completamente reestructurada por los avances de la tecnología digital, y hoy en día tenemos acceso a miles de canciones almacenadas en nuestros dispositivos móviles y a millones más a través de servicios en la nube. Dada esta inmensa cantidad de música al nuestro alcance, necesitamos nuevas maneras de describir, indexar, buscar e interactuar con el contenido musical. Esta tesis se centra en una tecnología que abre las puertas a nuevas aplicaciones en este área: la extracción automática de la melodía a partir de una grabación musical polifónica. Mientras que identificar la melodía de una pieza es algo que los humanos pueden hacer relativamente bien, hacerlo de forma automática presenta mucha complejidad, ya que requiere combinar conocimiento de procesado de señal, acústica, aprendizaje automático y percepción sonora. Esta tarea se conoce en el ámbito de investigación como “extracción de melodía”, y consiste técnicamente en estimar la secuencia de alturas correspondiente a la melodía predominante de una pieza musical a partir del análisis de la señal de audio. Esta tesis presenta un método innovador para la extracción de la melodía basado en el seguimiento y caracterización de contornos tonales. En la tesis, mostramos cómo se pueden explotar las características de contornos en combinación con reglas basadas en la percepción auditiva, para identificar la melodía a partir de todo el contenido tonal de una grabación, tanto de manera heurística como a través de modelos aprendidos automáticamente. A través de una iniciativa internacional de evaluación comparativa de algoritmos, comprobamos además que el método propuesto obtiene resultados punteros. De hecho, logra la precisión más alta de todos los algoritmos que han participado en la iniciativa hasta la fecha. Además, la tesis demuestra la utilidad de nuestro método en diversas aplicaciones tanto de investigación como para usuarios finales, desarrollando una serie de sistemas que aprovechan la melodía extraída para la búsqueda de música por semejanza (identificación de versiones y búsqueda por tarareo), la clasificación del estilo musical, la transcripción o conversión de audio a partitura, y el análisis musical con métodos computacionales. La tesis también incluye un amplio análisis comparativo del estado de la cuestión en extracción de melodía y el primer análisis crítico existente de la metodología de evaluación de algoritmos de este tipo
La indústria musical va ser una de les primeres a veure's completament reestructurada pels avenços de la tecnologia digital, i avui en dia tenim accés a milers de cançons emmagatzemades als nostres dispositius mòbils i a milions més a través de serveis en xarxa. Al tenir aquesta immensa quantitat de música al nostre abast, necessitem noves maneres de descriure, indexar, buscar i interactuar amb el contingut musical. Aquesta tesi es centra en una tecnologia que obre les portes a noves aplicacions en aquesta àrea: l'extracció automàtica de la melodia a partir d'una gravació musical polifònica. Tot i que identificar la melodia d'una peça és quelcom que els humans podem fer relativament fàcilment, fer-ho de forma automàtica presenta una alta complexitat, ja que requereix combinar coneixement de processament del senyal, acústica, aprenentatge automàtic i percepció sonora. Aquesta tasca es coneix dins de l'àmbit d'investigació com a “extracció de melodia”, i consisteix tècnicament a estimar la seqüència de altures tonals corresponents a la melodia predominant d'una peça musical a partir de l'anàlisi del senyal d'àudio. Aquesta tesi presenta un mètode innovador per a l'extracció de la melodia basat en el seguiment i caracterització de contorns tonals. Per a fer-ho, mostrem com es poden explotar les característiques de contorns combinades amb regles basades en la percepció auditiva per a identificar la melodia a partir de tot el contingut tonal d'una gravació, tant de manera heurística com a través de models apresos automàticament. A més d'això, comprovem a través d'una iniciativa internacional d'avaluació comparativa d'algoritmes que el mètode proposat obté resultats punters. De fet, obté la precisió més alta de tots els algoritmes proposats fins la data d'avui. A demés, la tesi demostra la utilitat del mètode en diverses aplicacions tant d'investigació com per a usuaris finals, desenvolupant una sèrie de sistemes que aprofiten la melodia extreta per a la cerca de música per semblança (identificació de versions i cerca per taral•larà), la classificació de l'estil musical, la transcripció o conversió d'àudio a partitura, i l'anàlisi musical amb mètodes computacionals. La tesi també inclou una àmplia anàlisi comparativa de l'estat de l'art en extracció de melodia i la primera anàlisi crítica existent de la metodologia d'avaluació d'algoritmes d'aquesta mena.

博士
國立成功大學
資訊工程學系碩博士班
95
In recent year, the information retrieval systems obtain dramatically improvement to retrieve huge web documents precisely. While the increasing of multimedia documents, the multimedia information retrieval, including spoken document or broadcast news document retrieval, becomes more and more important but still remains many challenges. The aim of this dissertation is to develop an audio content analysis system to process the broadcast news audio stream in advance, so as to provide the audio change-point, audio type, content transcription, story topic of speech segments and speech summary for further application, such as spoken document retrieval. The approaches described in this dissertation include the audio segmentation and classification, noisy speech recognition, story segmentation and classification and speech summarization. First, an audio segmentation and classification approach are proposed for segmenting and classifying an audio stream based on audio type, including speech, music, speech with music background, speech with noise background and noise. A minimum description length (MDL)-based Gaussian model with multiple change-points window is constructed to statistically characterize the audio features. Audio segmentation segments the audio stream into a sequence of homogeneous sub-segments using the MDL-based Gaussian model via binary segmentation algorithm. Finally, a heuristic method is adopted to smooth the sub-segment sequence and provide the final segmentation and classification results via a segment-based audio classifier. After locating the change-points positions and identifying all the audio types, robust noisy speech recognition via feature enhancement is developed to transcribe the speech content since much speech segments uttered under noisy environment. Three prior models are introduced to characterize clean speech, noise and noisy speech respectively. Sequential noise estimation is therefore employed for prior model construction based on noise-normalized stochastic vector mapping. Furthermore, an environment model adaptation is also adopted to reduce the mismatch between training data and test data. After transcribing all the speech segments, a story segmentation and topic classification is proposed to locate the topic change-points in contiguous stories and to identify the topic for every speech segment. A two-stage paradigm adopts a decision tree and maximum entropy model to identify the potential story boundaries in the broadcast news within a sliding window is conducted first. The story segmentation problem for story segmentation is thus transformed to the determination of a boundary position sequence from the potential boundary regions. The genetic algorithm is then applied to determine the chromosome, which corresponds to the final boundary position sequence. A topic-based segmental model is proposed to define the fitness function applied in the genetic algorithm. Finally, a speech summarization is further involved. A word sequence which maximizes a summarization score, including speech recognition confidence, prosody information, word significance, word trigram and semantic dependency relation, is extracted from automatically transcribed speech. Second, in the spoken documents, all the speech segments corresponding to the words in the summarized word sequence are extracted as the candidates for concatenation. Finally, the speech segments with the highest concatenation score among the candidates are selected and concatenated to generate a smooth summarized speech output. This speech summarization function not only keeps the most important information of every story but provides a concise speech summary representation, which is more friendly and time-saving representation for further spoken document retrieval and browsing.

碩士
中華大學
資訊工程學系(所)
96
From the 1990s until now, the software and hardware of computer on the growth rates have reached an unprecedented height. On the one hand, with a lot of ultra large databases to build, such as retrieval, search and classification of the metadata, is an importance topic of research that can't be ignored. In digital multimedia, the storing formats and function of various would let the digital audio-visual can be constantly evolved to becomes more and more perfect through the ISO MPEG standard. In the evolution of audio formats , it is stereo to reach from the mono , and then reach the multi-channel format . Focusing on the AAC(Advanced Audio Coding), it is a multi-channel format that ISO replaces the stereo format and establishes the audio kernel of MPEG-4 standard. With the rapid development of DVB(Digital Video Broadcasting), that produced a lot of researches relative to the subject. Based on all kinds of audio category correspond with a variety of TV commercials, this subject is rich that relation would effect the commercial analysis of the multimedia or database. Under the audio specification of DVB-T, we will introduce and analyze about this format basis for the AAC of DVB background; meanwhile, we also present a segment method to make a sequence of break-point detection, which is segment the content of commercial by use of the feature differentia of Euclidean distance. On the content classification of commercial, we modified a classic model of the vector space to build the classification model of commercial audio, and then identify every kind of audio category through the similarity comparison of feature path of TF-IDF(Term Frequency-Inverse Document Frequency) weight. Under the foundation of the audio-feature descriptor of MPEG-7, the automation technology of AAC analysis would be make the content analysis of automation with the commercial audio.

The goal of this research project is to undertake a critical evaluation of signal representations for musical audio content analysis. In particular it will contrast three different means for undertaking the analysis of micro-rhythmic content in Afro-Latin American music, namely through the use of: i) stereo or mono mixed recordings; ii) separated sources obtained via state of the art musical audio source separation techniques; and iii) the use of perfectly separated multi-track stems. In total the project comprises the following four objectives: i) To compile a dataset of mixed and multi-channel recordings of the Brazilian Maracatu musicians; ii) To conceive methods for rhythmical micro-variations analysis and pattern recognition; iii) To explore diverse music source separation approaches that preserve micro-rhythmic content; iv) To evaluate the performance of several automatic onset estimation approaches; and v) To compare the rhythmic analysis obtained from the original multi-channel sources versus the separated ones to evaluate separation quality regarding microtiming identification.

The goal of this research project is to undertake a critical evaluation of signal representations for musical audio content analysis. In particular it will contrast three different means for undertaking the analysis of micro-rhythmic content in Afro-Latin American music, namely through the use of: i) stereo or mono mixed recordings; ii) separated sources obtained via state of the art musical audio source separation techniques; and iii) the use of perfectly separated multi-track stems. In total the project comprises the following four objectives: i) To compile a dataset of mixed and multi-channel recordings of the Brazilian Maracatu musicians; ii) To conceive methods for rhythmical micro-variations analysis and pattern recognition; iii) To explore diverse music source separation approaches that preserve micro-rhythmic content; iv) To evaluate the performance of several automatic onset estimation approaches; and v) To compare the rhythmic analysis obtained from the original multi-channel sources versus the separated ones to evaluate separation quality regarding microtiming identification.

The abundance and ubiquity of user-generated content has opened horizons when it comes to the organization and analysis of vast and heterogeneous data, especially with the increase of quality of the recording devices witnessed nowadays. Most of the activity experienced in social networks today contains audio excerpts, either by belonging to a certain video file or an actual audio clip, therefore the analysis of the audio features present in such content is of extreme importance in order to better understand it. Such understanding would lead to a better handling of ubiquity data and would ultimately provide a better experience to the end-user. The work discussed in this thesis revolves around using audio features to organize and retrieve meaningful insights from user-generated content crawled from social media websites, more particularly data related to concert clips. From its redundancy and abundance (i.e., for the existence of several recordings of a given event), recordings from musical shows represent a very good use case to derive useful and practical conclusions around the scope of this thesis. Mechanisms that provide a better understanding of such content are presented and already partly implemented, such as audio clustering based on the existence of overlapping audio segments between different audio clips, audio segmentation that synchronizes and relates the different cluster’s clips in time, and techniques to infer audio quality of such clips. All the proposed methods use information retrieved from an audio fingerprinting algorithm, used for the synchronization of the different audio files, with methods for filtering possible false positives of the algorithm being also presented. For the evaluation and validation of the proposed methods, we used one dataset made of several audio recordings regarding different concert clips manually crawled from YouTube.

no
Purpose – Audio-diary methods are under-utilised in contemporary qualitative research. The purpose of this paper is to discuss participants and researchers’ experiences of using audio-diaries alongside semi-structured interviews to explore breastfeeding experiences in a short-term longitudinal study with 22 first-time mothers. Design/methodology/approach – The authors provide a qualitative content analysis of the participants’ feedback about their experiences of the audio-diary method and supplement this with the perspectives of the research team based on fieldwork notes, memos and team discussions. The authors pay particular attention to the ways in which the data attained from diaries compared with those from the interviews. Findings – The diaries produced were highly heterogeneous in terms of data length and quality. Participants’ experiences with the method were varied. Some found the process therapeutic and useful for reflecting upon the development of breastfeeding skills whilst negative aspects related to lack of mobility, self-consciousness and concerns about confidentiality. Researchers were positive about the audio-diary method but raised certain ethical, epistemological and methodological concerns. These include debates around the use of prompts, appropriate support for participants and the potential of the method to influence the behaviour under scrutiny. Interview and diary accounts contrasted and complemented in ways which typically enriched data analysis. Practical implications – The authors conclude that audio-diaries are a flexible and useful tool for qualitative research especially within critical realist and phenomenological paradigms. Originality/value – This appears to be the first paper to evaluate both participants and researchers’ experiences of using audio-diaries in a detailed and systematic fashion.

碩士
國立臺灣師範大學
資訊工程研究所
100
Most of the classrooms come with blackboards, and blackboards are widely used as a teaching prop in lecture video recordings. However, there are very few discussions about lecture video recordings that use blackboard as teaching prop concerning its multimedia semantics analysis. The article used a visual and optical based research method to explore speaker’s body languages and tone of speech in the blackboard lecture recordings, and how the amount of attention to pay in different segments of lecture recordings to enhance students’ learning. The visual analysis focused on semantics implied in speaker’s postures. The optical analysis focused on the variations of speaker’s speech emotions in his flow of teaching. The article proposed a speech emotion recognition model that divides speech emotions into five categories of happy, angry, bored, sad, and normal. The results of the analysis showed semantic intensity of the speaker and the importance of speakers teaching in different segments, and how students can learn more effectively with their variations in amount of attention according to the importance of speakers’ teaching throughout lecture video recordings.

碩士
國立交通大學
科技管理研究所
106
Due to the continued expansion of the music market and the popularity of streaming music services, it's becoming increasingly important to highlight personalized recommendation systems. And the music similarity measures in content-based recommendation systems is often incorporated into the mainstream recommendations system to provide personalized music recommendations. In the context of the world's largest music streaming service provider: Spotify released audio features API, there is no literature to explain the audio features and to propose a similarity measuring method for it. This study collects the audio feature data of users' favorite songs through the Spotify API, proposes novel similarity measure method. We discussed and compared the algorithms to find out the best algorithms for each step of similarity measure method (including Gap Statistic to find the optimal number of clusters, K-means clustering, machine learning algorithm LightGBM to find the feature importance, feature weighting, Euclidean distance to calculate music similarity, etc.). Furthermore, we discussed each audio feature provided by the Spotify API. The results show that the similarity measuring method have a considerable degree of credibility and can be extended to construct the current mainstream recommendation system. In addition, due to the high importance of customer preference structure for personalized recommendation, this study analyzed the novelty and diversity of user preference and divided the users into four groups based on their novelty and diversity. In the results, we combined the similarity measuring method and the preferences of users to give the recommended strategies to each group. In the future research, we will increase the size of the dataset, add more variables (ex. regional variables, personal variables, etc.), apply our method on another kind of audio data such as MFCCs, and extend the similarity measuring method and preference structure to other fields of research (ex. Shopping, Movies, Economic, etc.).

abstract: Audio signals, such as speech and ambient sounds convey rich information pertaining to a user’s activity, mood or intent. Enabling machines to understand this contextual information is necessary to bridge the gap in human-machine interaction. This is challenging due to its subjective nature, hence, requiring sophisticated techniques. This dissertation presents a set of computational methods, that generalize well across different conditions, for speech-based applications involving emotion recognition and keyword detection, and ambient sounds-based applications such as lifelogging. The expression and perception of emotions varies across speakers and cultures, thus, determining features and classification methods that generalize well to different conditions is strongly desired. A latent topic models-based method is proposed to learn supra-segmental features from low-level acoustic descriptors. The derived features outperform state-of-the-art approaches over multiple databases. Cross-corpus studies are conducted to determine the ability of these features to generalize well across different databases. The proposed method is also applied to derive features from facial expressions; a multi-modal fusion overcomes the deficiencies of a speech only approach and further improves the recognition performance. Besides affecting the acoustic properties of speech, emotions have a strong influence over speech articulation kinematics. A learning approach, which constrains a classifier trained over acoustic descriptors, to also model articulatory data is proposed here. This method requires articulatory information only during the training stage, thus overcoming the challenges inherent to large-scale data collection, while simultaneously exploiting the correlations between articulation kinematics and acoustic descriptors to improve the accuracy of emotion recognition systems. Identifying context from ambient sounds in a lifelogging scenario requires feature extraction, segmentation and annotation techniques capable of efficiently handling long duration audio recordings; a complete framework for such applications is presented. The performance is evaluated on real world data and accompanied by a prototypical Android-based user interface. The proposed methods are also assessed in terms of computation and implementation complexity. Software and field programmable gate array based implementations are considered for emotion recognition, while virtual platforms are used to model the complexities of lifelogging. The derived metrics are used to determine the feasibility of these methods for applications requiring real-time capabilities and low power consumption.
Dissertation/Thesis
Doctoral Dissertation Electrical Engineering 2015

Yes
Despite the importance of the ongoing debate on audit tendering and its possible implications for the audit profession including audit market structure, audit quality, and auditor independence, there is an apparent lack of research into this area. Using content analysis, this study reports the results of an examination of the comment letters sent to the UK Financial Reporting Council (FRC) in response to its consultation document on the 2012 revisions of the UK Corporate Governance Code. The results indicate a general support for the FRC’s proposals with a number of key concerns related to audit quality, audit cost and auditor independence. There is also clear conflict of interests among some groups such as audit firms and companies on one side and institutional investors on the other side. There is evidence of conflict of interest between Big 4 and non-Big 4 audit firms. The findings could influence future revisions of the Code with regard to tendering and enhance policy makers’ understanding of the position taken by each group of stakeholder.

碩士
國立高雄應用科技大學
會計系
106
This study conducted a content mining technique to analyze the articles based on bibliographic coupling and co-word analyses to determine the salient research subjects. These audit quality research subjects were verified with those proxies by the International Auditing and Assurance Standard Boards (IAASB) framework with a total 520 articles from 1992 to 2017 were analyzed. The result found 14 topics based on bibliographic coupling and 17 topics based co-word analyses. In terms of most contributing countries, USA become the most influential countries in audit quality research both of two methods. Moreover, this study also identified the highly cited references and most productive authors during 1992 to 2017. The results showed that auditor characteristic (AC) was the most attractive topics in audit quality research during 25 years based on BC analysis, differently auditor-client contracting features (ACCF) leading the highest discussed topics from 1992 to 2017 based on co-word analysis. Regarding to Social Network Analysis (SNA) result, Contextual Factor element as the main issue in BC analysis, while Input element located in the very center of the whole co-word analysis network. The results of this study will help researchers look into their own field specialties, understand the developing trends in audit quality research, indicate researchers who have had influence in audit quality research, and determine future research subjects.

Text in Afrikaans
Hierdie studie is onderneem om aan te dui hoe die verslagleweringsfunksie van die ouditeur in openbare praktyk be1nvloed word deur die voortdurende veranderings in onder andere, gemeenskapstrukture en gemeenskapsdoelwitte. Gedurende die afgelope aantal jare het die gebruiker van die ouditeursversl ag herhaaldelik druk op die ouditprofessie geplaas om beter kommunikasiemiddels tot stand te bring. Die ouditprofessie het gehoor hier aan gegee en een van die reaksies op hierdie versoek was die uitreiking van 'n voorgeskrewe nuwe verlengde vorm van die ouditeursverslag. In hierdie studie word die proses van ontwikkeling van die verlengde vorm van die ouditeursverslag deurgevolg. Spesifieke verwysing na die verslagleweringsbeleide wat tans in sekere toonaangewende lande gevolg word, word gedoen, naamlik: die Verenigde State van Amerika, Suid-Afrika en die Verenigde Koninkryk. Om te verseker dat die ouditprofessie altyd gehoor gee aan die behoeftes van die gebruiker van die ouditeursverslag en sodoende 'n professie met 'n visie sal bly, word aanbeveel dat volgehoue navorsing op die gebied van ouditverslaglewering sal bly voortgaan.
This study was undertaken to indicate how continua 1 changes in, inter alia, community structures and goals have influenced the reporting function of the auditor in public practice. In recent years users of the auditor's report have applied continual pressure on the profession to improve its communication tools. One of the responses to this demand has been the issuing of the newly expanded long format auditor's report. In this study the development process of the new long format auditor's report has been researched and described with specific reference to the reporting practices in some of the leading countries in the field viz; the United States of America,South Africa and the United Kingdom. Continuing research within the domain of audit reporting is recommended to ensure that the profession deals with reporting problems as they occur, thus retaining the visionary status of the profession.
Auditing
M. Com. (Auditing)

Yes
Background: Interpreting evaluations of complex interventions can be difficult without sufficient description of key intervention content. We aimed to develop an implementation package for primary care which could be delivered using typically available resources and could be adapted to target determinants of behaviour for each of four quality indicators: diabetes control, blood pressure control, anticoagulation for atrial fibrillation and risky prescribing. We describe the development and prospective verification of behaviour change techniques (BCTs) embedded within the adaptable implementation packages. Methods: We used an over-lapping multi-staged process. We identified evidence-based, candidate delivery mechanisms—mainly audit and feedback, educational outreach and computerised prompts and reminders. We drew upon interviews with primary care professionals using the Theoretical Domains Framework to explore likely determinants of adherence to quality indicators. We linked determinants to candidate BCTs. With input from stakeholder panels, we prioritised likely determinants and intervention content prior to piloting the implementation packages. Our content analysis assessed the extent to which embedded BCTs could be identified within the packages and compared them across the delivery mechanisms and four quality indicators. Results: Each implementation package included at least 27 out of 30 potentially applicable BCTs representing 15 of 16 BCT categories. Whilst 23 BCTs were shared across all four implementation packages (e.g. BCTs relating to feedback and comparing behaviour), some BCTs were unique to certain delivery mechanisms (e.g. ‘graded tasks’ and ‘problem solving’ for educational outreach). BCTs addressing the determinants ‘environmental context’ and ‘social and professional roles’ (e.g. ‘restructuring the social and ‘physical environment’ and ‘adding objects to the environment’) were indicator specific. We found it challenging to operationalise BCTs targeting ‘environmental context’, ‘social influences’ and ‘social and professional roles’ within our chosen delivery mechanisms. Conclusion: We have demonstrated a transparent process for selecting, operationalising and verifying the BCT content in implementation packages adapted to target four quality indicators in primary care. There was considerable overlap in BCTs identified across the four indicators suggesting core BCTs can be embedded and verified within delivery mechanisms commonly available to primary care. Whilst feedback reports can include a wide range of BCTs, computerised prompts can deliver BCTs at the time of decision making, and educational outreach can allow for flexibility and individual tailoring in delivery
UK National Institute for Health Research (NIHR) under its Programme Grants for Applied Research scheme (Grant Reference Number RP-PG-1209-10040).