This dissertation features methods of analyzing symbolic music, focused on n-gram-based approaches, as this representation resembles the most text and natural languages. The analysis of similarities between several text and music corpora is accompanied with implementation of text-based methods for problems of composer classification and symbolic music similarity definition. Both problems contain thorough evaluation of performance of the systems with comparisons to other approaches on existing testbeds. It is also described how one can use this symbolic representation in conjunction with genetic algorithms to tackle problems like melody generation. The proposed method is fully automated, and the process utilizes n-gram statistics from a sample corpus to achieve it. A method of visualization of complex symbolic music pieces is also presented. It consist of creating a self similarity matrix of a piece in question, revealing dependencies between voices, themes and sections, as well as music structure. A fully automatic technique of inferring music structure from these similarity matrices is also presented The proposed structure analysis system is compared against similar approaches that operate on audio data. The evaluation shows that the presented structure analysis system outperformed significantly all audio-based algorithms available for comparison in both precision and recall.

The visualization of the high-dimensional feature landscapes that are encountered when analyzing audio data is a challenging problem and is the focus of much research in the field of Music Information Retrieval. Typical feature sets extracted from sound have anywhere from dozens to hundreds of dimensions and have complex interrelationships between data elements. In this work, we apply various modern techniques for the visualization of audio data to a number of diverse problem domains, including the bioacoustics of Orcinus Orca (killer whale) song, partially annotated chant traditions including Torah recitation and the the analysis of music collections and live DJ sets. We also develop a number of graphical user interfaces to allow users to interact with these visualizations. These interfaces include Flash-enabled web applications, desktop applications, and novel interfaces including the use of the Radiodrum, a three-dimension position sensing musical interface.

Several content-based interfaces for music collection exploration rely on Self-Organizing Maps (SOMs) to produce 2D or 3D visualizations of music spaces. In these visualizations, perceptually similar songs are clustered together. The positions of clusters containing similar songs, however, cannot be determined in advance due to particularities of the traditional SOM algorithm. In this thesis, I propose a variation on the traditional algorithm named anchoredSOM. This variation avoids changes in the positions of the aforementioned clusters. Moreover, anchoredSOM allows users to personalize the music space by choosing the locations of clusters containing per- ceptually similar tracks. This thesis introduces SoundAnchoring, an interface for music collection exploration featuring anchoredSOM. SoundAnchoring is evaluated by means of a user study. Results show that SoundAnchoring offers engaging ways to explore music collections and build playlists. / Graduate / 0984 / 0413 / leandro.collares@gmail.com

Acrescente capacidade de armazenamento introduziu novos desafios no contexto de exploração de grandes bases de dados de músicas. Esse trabalho consiste em investigar técnicas de comparação de músicas representadas por sinais polifônicos, com o objetivo de encontrar similaridades, permitindo a identificação de músicas cover em grandes bases de dados. Técnicas de extração de características a partir de sinais musicais foram estudas, como também métricas de comparação a partir das características obtidas. Os resultados mostraram que é possível encontrar um novo método de identificação de covers com um menor custo computacional do que os existentes, mantendo uma boa precisão / The growing capacity in storage and transmission of songs has introduced a new challenges in the context of large music data sets exploration. This work aims at investigating techniques for comparison of songs represented by polyphonic signals, towards identifying cover songs in large data sets. Techniques for music feature extraction were evaluated and compared. The results show that it is possible to develop new methods for cover identification with a lower computational cost when compared to existing solutions, while keeping the good precision

Submitted by Rafael Santana (rafael.silvasantana@ufpe.br) on 2017-08-31T18:49:43Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Dissertação Renato.pdf: 2458758 bytes, checksum: b08fb4f41c821e5fd07c0022ea5dcaac (MD5) / Made available in DSpace on 2017-08-31T18:49:43Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Dissertação Renato.pdf: 2458758 bytes, checksum: b08fb4f41c821e5fd07c0022ea5dcaac (MD5) Previous issue date: 2016-09-14 / Em aplicações de Preview de serviços de streaming de música, onde uma rápida impressão de um álbum desconhecido é proporcionada pela navegação de suas músicas, a inclusão do refrão no trecho de trinta segundos fornecido para cada música torna a aplicação muito mais precisa e eficaz. O refrão pode também funcionar como uma “miniatura” representativa da música, melhorando o desempenho e a precisão das consultas, se realizadas somente procurando pelos refrãos em vez de se procurar por músicas inteiras. Diante da importância de obter o trecho mais representativo de uma canção, o objetivo de um sistema de detecção de refrão é identificar este segmento ou, mais precisamente, os seus instantes inicial e final. Métodos do Estado da Arte buscam extrair features associadas a notas musicais e timbre como vetores Chroma e MFCC, e a partir destas identificar as repetições entre os segmentos da música, inclusive o refrão. Este tipo de abordagem torna o método pouco robusto no processamento de músicas onde notas musicais e variedade de timbres não são tão presentes, como em estilos musicais mais percussivos. Este trabalho propõe uma mudança de paradigma para a detecção de refrão, baseada na exploração do domínio do tempo em lugar do domínio da frequência, com o objetivo de obter um método mais competitivo no processamento de músicas percussivas. O método proposto elimina a etapa de segmentação, substitui as features harmônicas e timbrais pela envoltória do sinal e utiliza a função de correlação entre as envoltórias das partes da música como métrica de similaridade, tornando o método menos dependente de notas musicais e timbres. Os testes mediram o grau de degeneração das taxas de acertos do método proposto e de uma versão modificada usando vetores de Chroma sobre uma base harmônica e uma base percussiva. Os resultados indicam que a abordagem proposta sofre uma degeneração duas vezes menor que a versão modificada, comprovando a hipótese de que um método de detecção de refrão que explore o domínio do tempo é mais competitivo, ao processar músicas percussivas, que um método limitado à exploração do domínio da frequência. / In Preview applications of music streaming services, where a fast printing from an unknown album is provided by the navigation of your songs, including the chorus in thirty seconds excerpt provided for each song makes the application much more accurate and effective. The chorus can also function as a “miniature” representative of music, enhancing the performance and accuracy of search, if carried out only by looking choruses instead of searching for entire songs. Given the importance of getting the most representative excerpt of a song, the goal of a chorus detection system is to identify this segment, or more precisely, its beginning and its end. State of the art methods seek to extract features associated with musical notes and timbre, like Chroma and MFCC vectors and identify from these repetitions between segments of music, including the chorus. This approach type makes method little robust in music where musical notes and variety of timbres are not as present, as in percussive music for example. This paper proposes a paradigm shift for the chorus detection, based on the exploitation of the time domain instead of the frequency domain, in order to obtain a more competitive method in the processing of percussive music. The proposed method eliminates the segmentation, replaces the harmonic and timbral features with the envelope of the signal, and uses the correlation function between the envelope of the music segments as a metric of similarity, to make it less dependent on musical notes and timbre. The tests measured the degree of degeneration of hit rates of the proposed method and of a modified version using Chroma vectors on a harmonic basis and a percussive basis. The results indicate that the proposed approach have a degeneration two times lower than the modified version, proving the hypothesis that a chorus detection method that exploits the time domain is more competitive when processing percussive songs than a method limited to the frequency domain exploitation.

La transcripción de música por computador es de vital importancia en tareas del llamo campo de la Extracción y recuperación de información musical por su utilidad como proceso para la obtención de una abstracción simbólica que codifica el contenido musical de un fichero de audio. En esta disertación se estudia este problema desde una perspectiva diferente a la típicamente considerada para estos problemas, la perspectiva interactiva y multimodal. En este paradigma el usuario cobra especial importancia puesto que es parte activa en la resolución del problema (interactividad); por otro lado, la multimodalidad implica que diferentes fuentes de información extraídas de la misma señal se aúnan para ayudar a una mejor resolución de la tarea.

Cette thèse s'inscrit dans le cadre de l'analyse automatique de la musique.La finalité de ce champ de recherche est d'extraire des informations de la musique, ou autrement dit, de faire comprendre ce qu'est la musique à un ordinateur.Les applications sont nombreuses: fabriquer des systèmes de recommandation musicale, transcrire une partition à partir du signal ou générer automatiquement de la musique.Nous nous intéressons dans ce manuscrit à l'analyse automatique du rythme.Notre objectif est de proposer de nouvelles descriptions du rythme qui s'inspirent d'études perceptives et neurologiques.La représentation du rythme d’un signal musical audio est un problème complexe.Il ne s’agit pas simplement de détecter la position des attaques et la durée des notes comme sur une partition mais plus généralement de modéliser l’interaction temporelle entre les différents instruments présents et collaborant à l’établissement d’un rythme de manière compacte, discriminante et invariante.Nous cherchons à obtenir des représentations invariantes à certains paramètres (tels la position dans le temps, les variations faibles de tempo ou d’instrumentation) mais à l’inverse sensibles à d’autres (comme le motif rythmique, les paramètres fins d’interprétation ou le swing). Nous étudions les trois aspects fondamentaux pour la description du rythme: le tempo les déviations et les motifs rythmiques. / This thesis is within the scope of Music Information Retrieval. The goal of this research field is to extract meaningful informations from music. There are numerous applications: music recommendation systems, music transcription to a score or automatic generation of music. In this manuscript, oOur objective is to propose new rhythm descriptions inspired from perceptual and neurological studies.Rhythm representation of a musical signal is a complex problem. Detecting attack positions and note durations is not sufficient: we have the model the temporal interaction between the different instruments collaborating together to create rhythm. We try to obtain representations that are invariant to some parameters (like the position over time, the small tempo or instrumentation variations) but sensitive to other parameters (like the rhythm pattern or the swing factor). We study the three key aspect of rhythm description: tempo, deviations and rhythm pattern.

Les tags constituent un outil très utile pour indexer des documents multimédias. Cette thèse de doctorat s’intéresse au tagging automatique, c’est à dire l’association automatique par un algorithme d’un ensemble de tags à chaque morceau. Nous utilisons des techniques de boosting pour réaliser un apprentissage prenant mieux en compte la richesse de l’information exprimée par la musique. Un algorithme de boosting est proposé, afin d’utiliser conjointement des descriptions de morceaux associées à des extraits de différentes durées. Nous utilisons cet algorithme pour fusionner de nouvelles descriptions, appartenant à différents niveaux d’abstraction. Enfin, un nouveau cadre d’apprentissage est proposé pour le tagging automatique, qui prend mieux en compte les subtilités des associations entre les tags et les morceaux. / Tags constitute a very useful tool for multimedia document indexing. This PhD thesis deals with automatic tagging, which consists in associating a set of tags to each song automatically, using an algorithm. We use boosting techniques to design a learning which better considers the complexity of the information expressed by music. A boosting algorithm is proposed, which can jointly use song descriptions associated to excerpts of different durations. This algorithm is used to fuse new descriptions, which belong to different abstraction levels. Finally, a new learning framework is proposed for automatic tagging, which better leverages the subtlety ofthe information expressed by music.

La modélisation computationnelle du rythme a pour objet l'extraction et le traitement d’informations rythmiques à partir d’un signal audio de musique. Cela s'avère être une tâche extrêmement complexe car, pour traiter un enregistrement audio réel, il faut pouvoir gérer sa complexité acoustique et sémantique à plusieurs niveaux de représentation. Les méthodes d’analyse rythmique existantes se concentrent généralement sur l'un de ces aspects à la fois et n’exploitent pas la richesse de la structure musicale, ce qui compromet la cohérence musicale des estimations automatiques. Dans ce travail, nous proposons de nouvelles approches tirant parti des informations multi-échelles pour l'analyse automatique du rythme. Nos modèles prennent en compte des interdépendances intrinsèques aux signaux audio de musique, en permettant ainsi l’interaction entre différentes échelles de temps et en assurant la cohérence musicale entre elles. En particulier, nous effectuons une analyse systématique des systèmes de l’état de l’art pour la détection des premiers temps, ce qui nous conduit à nous tourner vers des architectures convolutionnelles et récurrentes qui exploitent la modélisation acoustique à court et long terme; nous introduisons un modèle de champ aléatoire conditionnel à saut de chaîne pour la détection des premiers temps. Ce système est conçu pour tirer parti des informations de structure musicale (c'est-à-dire des répétitions de sections musicales) dans un cadre unifié. Nous proposons également un modèle linguistique pour la détection conjointe des temps et du micro-timing dans la musique afro-latino-américaine. Nos méthodes sont systématiquement évaluées sur diverses bases de données, allant de la musique occidentale à des genres plus spécifiques culturellement, et comparés à des systèmes de l’état de l’art, ainsi qu’à des variantes plus simples. Les résultats globaux montrent que nos modèles d’estimation des premiers temps sont aussi performants que ceux de l’état de l'art, tout en étant plus cohérents sur le plan musical. De plus, notre modèle d’estimation conjointe des temps et du microtiming représente une avancée vers des systèmes plus interprétables. Les méthodes présentées ici offrent des alternatives nouvelles et plus holistiques pour l'analyse numérique du rythme, ouvrant des perspectives vers une analyse automatique plus complète de la musique. / Computational rhythm analysis deals with extracting and processing meaningful rhythmical information from musical audio. It proves to be a highly complex task, since dealing with real audio recordings requires the ability to handle its acoustic and semantic complexity at multiple levels of representation. Existing methods for rhythmic analysis typically focus on one of those levels, failing to exploit music’s rich structure and compromising the musical consistency of automatic estimations. In this work, we propose novel approaches for leveraging multi-scale information for computational rhythm analysis. Our models account for interrelated dependencies that musical audio naturally conveys, allowing the interplay between different time scales and accounting for music coherence across them. In particular, we conduct a systematic analysis of downbeat tracking systems, leading to convolutional-recurrent architectures that exploit short and long term acoustic modeling; we introduce a skip-chain conditional random field model for downbeat tracking designed to take advantage of music structure information (i.e. music sections repetitions) in a unified framework; and we propose a language model for joint tracking of beats and micro-timing in Afro-Latin American music. Our methods are systematically evaluated on a diverse group of datasets, ranging from Western music to more culturally specific genres, and compared to state-of-the-art systems and simpler variations. The overall results show that our models for downbeat tracking perform on par with the state of the art, while being more musically consistent. Moreover, our model for the joint estimation of beats and microtiming takes further steps towards more interpretable systems. The methods presented here offer novel and more holistic alternatives for computational rhythm analysis, towards a more comprehensive automatic analysis of music.

The ever-increasing density of computer storage devices has allowed the average user to store enormous quantities of multimedia content, and a large amount of this content is usually music. Current search techniques for musical content rely on meta-data tags which describe artist, album, year, genre, etc. Query-by-content systems, however, allow users to search based upon the actual acoustical content of the songs. Recent systems have mainly depended upon textual representations of the queries and targets in order to apply common string-matching algorithms and are often confined to a single query style (e.g., humming). These methods also lose much of the information content of the song which limits the ways in which a user may search. We present a query-by-content system which supports querying in several styles using a Self-Organizing Map as its basis. The results from testing our system show that it performs better than random orderings and is, therefore, a viable option for musical query-by-content.