Music Information Retrieval

The Musical Avatar: a visualization of musical preferences by means of audio content description

by Anna Xambo

Haro, Martín; Xambó, Anna; Fuhrmann, Ferdinand; Bogdanov, Dmitry; Gómez, Emilia and Herrera, Perfecto (2010). In: 5th Audio Mostly Conference, 15-17 Sep 2010, Piteå, Sweden.

The music we like (i.e. our musical preferences) encodes and communicates key information about ourselves. Depicting... more The music we like (i.e. our musical preferences) encodes and communicates key information about ourselves. Depicting such preferences in a condensed and easily understandable way is very appealing, especially considering the current trends in social network communication. In this paper we propose a method to automatically generate, given a provided set of preferred music tracks, an iconic representation of a user's musical preferences - the Musical Avatar. Starting from the raw audio signal we first compute over 60 low-level audio features. Then, by applying pattern recognition methods, we infer a set of semantic descriptors for each track in the collection. Next, we summarize these track-level semantic descriptors, obtaining a user profile. Finally, we map this collection-wise description to the visual domain by creating a humanoid cartoony character that represents the user's musical preferences. We performed a proof-of-concept evaluation of the proposed method on 11 subjects with promising results. The analysis of the users' evaluations shows a clear preference for avatars generated by the proposed semantic descriptors over avatars derived from neutral or randomly generated values. We also found a general agreement on the representativeness of the users' musical preferences via the proposed visualization strategy.

Music Identification using Chroma Features

by Adrian Iftene

Audio Music Similarity and Retrieval: Evaluation Power and Stability

by Jorge Morato

J. Urbano, D. Martín, M. Marrero and J. Morato, “Audio Music Similarity and Retrieval: Evaluation Power and Stability“, 12th International Society for Music Information Retrieval Conference (ISMIR), October 24-28, 2011, Miami (USA), ), pp. 597-602

In this paper we analyze the reliability of the results in the evaluation of Audio Music Similarity and Retrieval... more In this paper we analyze the reliability of the results in the evaluation of Audio Music Similarity and Retrieval systems.
We focus on the power and stability of the evaluation, that is, how often a significant difference is found between systems and how often these significant differences are incorrect. We study the effect of using different effectiveness measures with different sets of relevance judgments, for varying number of queries and alternative statistical procedures. Different measures are shown to behave similarly overall, though some are much more sensitive and stable than others. The use of different statistical procedures does improve the reliability of the results, and it allows using as
little as half the number of queries currently used in MIREX
evaluations while still offering very similar reliability levels.
We also conclude that experimenters can be very confident
that if a significant difference is found between two systems,
the difference is indeed real.

Symbolic Melodic Similarity: Local Alignment with Geometric Representations

by Jorge Morato

Urbano, J., Llorens, J, Morato, J., and Sanchez-Cuadrado, S (2010) Symbolic Melodic Similarity: Local Alignment with Geometric Representations. Music Information Retrieval Evaluation eXchange MIREX 2010

This short paper describes four submissions to the Symbolic Melodic Similarity task of the MIREX 2010 edition. All... more This short paper describes four submissions to the Symbolic Melodic Similarity task of the MIREX 2010 edition. All four submissions rely on a local-alignment approach between sequences of n-grams, and they differ mainly on the substitution score between two n-grams. This score is based on a geometric representation that shapes musical pieces as curves in the pitch-time plane. One of the systems described ranked first for all ten effectiveness measures used and the other three ranked from second to fifth, depending on the measure.

Automatic Sound Identification based on Prosodic Listening

by Prof Eduardo R. Miranda

Proceedings of the 17th International Congress on Acoustics, Rome, Italy. Published by the International Commission for Acoustics - ICA

This paper presents a system for the identification of vocal and vocal-like sounds by means of a classificatory scheme... more This paper presents a system for the identification of vocal and vocal-like sounds by means of a classificatory scheme based upon their prosodic attributes.1 The system firstly segments examples of the sound classes in question and performs a number of analyses on these segments in order to determine their prosodic attributes. Then, it builds a classificatory scheme, which will be used to identify an unknown sound. For example, consider the identification of the language of utterances spoken in English, French and Japanese. In this case, the system establishes a classificatory scheme from samples of speech in these three languages. Then, the system infers the language of unknown utterances by matching their prosodic information with the classificatory scheme. The system is not, however, limited to language identification tasks; it also can identify singing styles and a number of animal calls. Given the appropriate training data, the system is able to identify such sounds with great accuracy.

Creative Process in Free Improvisation

by José Menezes

This research investigates the creative and communicational processes used by improvisers in free improvised... more This research investigates the creative and communicational processes used by improvisers in free improvised performance and the ideologies behind those processes. Two studies were conducted: In Study 1 quantitative data was extracted from a recorded performance with Music Information Retrieval (M.I.R) software with special focus on moments consensually considered by the musicians as “best”. Study 2 analysed qualitative data extracted from interviews with improvisers and retrospective verbal protocol regarding the whole performance with special focus on “best moments”. The results of Study 1 reveal the use of alterations of musical features such as energy, note density and spectral changes in order to create points of qualitative change in improvised music. Creative strategies revealed by Study 2 include reiteration, the use of error as a motor for generation of music materials, real-time use of processes of musical composition and automatic playing. Improved conditions of separation of recorded instruments are advised in future research on this subject.

Everything in its right place? Learning a user's view of a music collection

by Korinna Bade

Supporting folk-song research by automatic metric learning and ranking

by Korinna Bade

Moody Blues: The Social Web, Tagging, and Nontextual Discovery Tools for Music

by Susannah Cleveland

co-authored with Gwen Evans

Music Reference Services Quarterly, Volume 11, Issue 3 & 4 December 2008, pages 177 - 201

A common... more Music Reference Services Quarterly, Volume 11, Issue 3 & 4 December 2008, pages 177 - 201

A common thread in discussions about the Next Generation Catalog is that it should incorporate features beyond the mere textual, one-way presentation of data. At the same time, traditional textual description of music materials often prohibits effective use of the catalog by specialists and nonspecialists alike. Librarians at Bowling Green State University have developed the HueTunes project to explore already established connections between music, color, and emotion, and incorporate those connections into a nontextual discovery tool that could enhance interdisciplinary as well as specialist use of the catalog.

MULTI-LABEL CLASSIFICATION OF MUSIC INTO EMOTIONS

by Konstantinos Trochidis

A., “Classification of Recorded Classical Music: A Methodology and a Comparative Study

by Ricardo Malheiro

Malheiro, R., Paiva, R., Mendes, A., Mendes, T. and Cardoso, A., “Classification of Recorded Classical Music: A Methodology and a Comparative Study”, in Proceedings of the First International ICSC Symposium on Brain Inspired Cognitive Systems, BICS’2004, Stirling, Scotland, August-2004 (Electronic Proceedings), ISBN: 1-85769-199-7

As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music... more As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music Distribution
industry. Consequently, tasks such as automatic music genre
classification address new and exciting research challenges.
Automatic music genre recognition involves is sues like
feature extraction and development of classifiers using  the
obtained features. 
We use the number of zero crossings, loudness, spectral
centroid, bandwidth and uniformity for feature extraction.
These features are statistically manipulated, making a total of
40 features. 
Regarding the task of genre modeling, we follow three
approaches: the  K-Nearest Neighbors (KNN) classifier,
Gaussian Mixture Models (GMM) and feedforward neural
networks (FFNN). 
A taxonomy of sub-genres of classical music is used. We
consider three classification problems: in the first one, we aim
at discriminating between music for flute, piano and violin; in
the second problem, we distinguish choral music from opera;
finally, in the third one, we seek to discriminate between al l
five genres.
The best results were obtained using FFNNs: 85%
classification accuracy in the three-class problem, 90% in the
two-class problem and 76% in the five-class problem. These
results are encouraging and show that the presented
methodology may be a good starting point for addressing more challenging tasks.

Sistemas de Classificação Musical com Redes Neuronais

by Ricardo Malheiro

Malheiro, R., Paiva, R., Mendes, A., Mendes, T. and Cardoso, A., “Sistemas de Classificação Musical com Redes Neuronais”, Revista Gestão e Desenvolvimento, Volume 12, pp. 167-195, Universidade Católica Portuguesa, Junho-2004, ISSN: 0872-556X.

Como resultado da evolução e inovação tecnológicas, a indústria da distribuição electrónica de música tem tido um... more Como resultado da evolução e inovação tecnológicas, a indústria da distribuição electrónica de música tem tido um enorme crescimento. Desta forma, tarefas como a classificação automática de géneros musicais tornam-se um fortíssimo motivo para o incremento da investigação na área. O reconhecimento automático de géneros musicais envolve tarefas como a extracção de características das músicas e o desenvolvimento de classificadores que utilizem essas características. Neste estudo pretendeu-se, através de 3 problemas de classificação independentes, classificar segmentos de música clássica em subgéneros. Para tal, foram extraídas 40 características por segmento musical, tendo os classificadores utilizados sido redes neuronais. Devido à qualidade dos resultados alcançados, foi construído em seguida um protótipo para um sistema real de classificação. Neste, de um conjunto de músicas não catalogadas, foram extraídos dez segmentos às “cegas” que foram classificados utilizando os classificadores anteriores. Cada música foi classificada no género mais representado pelos seus segmentos.

Classification of Recorded Classical Music using Neural Networks

by Ricardo Malheiro

Malheiro, R., Paiva, R., Mendes, A., Mendes, T. and Cardoso, A., "Classification of Recorded Classical Music using Neural Networks”, in Proceedings of the Fourth International ICSC Symposium on Engineering of Intelligent Systems, EIS'2004, Madeira, Portugal, February-2004 (Electronic Proceedings)

As a result of recent technological innovations, there has been a tremendous growth in the Electronic Music... more As a result of recent technological innovations, there has been a tremendous growth in the Electronic Music Distribution industry. In this way, tasks such us automatic music genre classification appear as new and exciting research challenges.
Automatic music genre recognition involves issues like feature extraction and development of classifiers using the obtained features.
As for feature extraction, we use the number of zero crossings, loudness, spectral centroid, bandwidth and uniformity. These features are statistically manipulated, making a total of 40 features.
Regarding the task of genre modeling, we train a feedforward neural network (FFNN) with the Levenberg Marquardt algorithm.
A taxonomy of subgenres of classical music is used. We consider three classification problems: in the first one, we aim to discriminate between music for flute, piano and violin; in the second problem, we distinguish choral music from opera; finally, in the third one, we aim to discriminate between all the abovementioned five genres together.
We obtained 85% classification accuracy in the three-class problem, 90% in the two-class problem and 76% in the five-class problem. These results are encouraging and show that the presented methodology may be a good starting point for addressing more challenging tasks.

A., “Classification of Recorded Classical Music: A Methodology and a Comparative Study

by Ricardo Malheiro

Malheiro, R., Paiva, R., Mendes, A., Mendes, T. and Cardoso, A., “Classification of Recorded Classical Music: A Methodology and a Comparative Study”, in Proceedings of the First International ICSC Symposium on Brain Inspired Cognitive Systems, BICS’2004, Stirling, Scotland, August-2004 (Electronic Proceedings), ISBN: 1-85769-199-7

As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music... more As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music Distribution
industry. Consequently, tasks such as automatic music genre
classification address new and exciting research challenges.
Automatic music genre recognition involves is sues like
feature extraction and development of classifiers using  the
obtained features. 
We use the number of zero crossings, loudness, spectral
centroid, bandwidth and uniformity for feature extraction.
These features are statistically manipulated, making a total of
40 features. 
Regarding the task of genre modeling, we follow three
approaches: the  K-Nearest Neighbors (KNN) classifier,
Gaussian Mixture Models (GMM) and feedforward neural
networks (FFNN). 
A taxonomy of sub-genres of classical music is used. We
consider three classification problems: in the first one, we aim
at discriminating between music for flute, piano and violin; in
the second problem, we distinguish choral music from opera;
finally, in the third one, we seek to discriminate between al l
five genres.
The best results were obtained using FFNNs: 85%
classification accuracy in the three-class problem, 90% in the
two-class problem and 76% in the five-class problem. These
results are encouraging and show that the presented
methodology may be a good starting point for addressing more challenging tasks.

A Prototype for Classification of Classical Music using Neural Networks

by Ricardo Malheiro

Malheiro, R., Paiva, R., Mendes, A., Mendes, T. and Cardoso, A., “A Prototype for Classification of Classical Music using Neural Networks”, in Proceedings of the Eighth IASTED International Conference on Artificial Intelligence and Soft Computing, pp. 294-299, ASC’2004, Marbella, Spain, September-2004 (Electronic Proceedings), ISSN: 1482-7913.

As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music... more As a result of recent technological innovations, there has
been a tremendous growth in the Electronic Music
Distribution industry. In this way, tasks such us automatic
music genre classification address new and exciting
research challenges. Automatic music genre recognition
involves issues like feature extraction and development of
classifiers using the obtained features. As for feature
extraction, we use features such as the number of zero
crossings, loudness, spectral centroid, bandwidth and
uniformity. These are statistically manipulated, making a
total of 40 features. As for the task of genre modeling, we
train a feedforward neural network (FFNN). A taxonomy
of subgenres of classical music is used. We consider three
classification problems: in  the first one, we aim at
discriminating between music for flute, piano and violin;
in the second problem, we distinguish choral music from
opera; finally, in the third one, we aim at discriminating
between all five genres. Preliminary results are presented
and discussed, which show that the presented
methodology may be a good starting point for addressing
more challenging tasks, such as using a broader range of
musical categories.

Segmentation and Discovery of Podcast Content

by Steven Hargreaves

"To Know Beyond Listening: Monitoring Digital Music"

by Ian Reyes

The Senses and Society (Fall, 2010)

In music production, “monitoring” refers traditionally to audile strategies intended to reveal the “true” sound of... more In music production, “monitoring” refers traditionally to audile strategies intended to reveal the “true” sound of mediated audio. Here, it is expanded to include new, digital technologies intended to better know and control the record-object beyond what listening and listening technologies allow. Surveying traditional, contemporary, and emerging tools of record production and distribution, this essay addresses three types of monitoring: audio, visual, and data.

In sum, monitoring entails the supplementation and subversion of the ear through protocols promising to surmount the biases and distortions of audio media. Key technologies include reference speakers, room correction systems, digital audio workstations, open mixes, pre-sets, social networking sites, and automatic music information retrieval. Situating these within a “techoustemology” of monitoring, the central argument is that many innovations in digital audio are non-auditory and, therefore, displace sound and listening as the central means of producing relevant knowledge about music mediated in the digital age.

EXPLOITING ONLINE RESOURCES TO IMPROVE CHORD RECOGNITION ACCURACY

by Tijl De Bie