| |
The ISMIR 2004 Graduate School took place from Monday to Saturday, 10/04 to 10/09 at the Auditorium, França area UPF. Here you can download the slides and presentations of the participants.
Chairs of the graduate school:
|
Xavier Serra
xserra@iua.upf.es
|
|
|
Mark Leman
marc.leman@ugent.be
|
|
Teachers:
Papers that professors asked to read before the courses
Students:
|
Andreas Finger
af@informatik.uni-rostock.de
Proposal
Implementation of content-based audio analysis in object-relational databases
|
|
|
Andrew Nesbit
andrew.nesbit@elec.qmul.ac.uk
Sound source separation for music remixing
|
|
| The topic is still in high-level stages and I am currently formalising the research questions to answer and how I can evaluate such a technique. Normally, sound source separation is concerned with extracting discrete sources or instruments from a mix, for example, vocals or drums. I want to find methods for decomposing a mix into musical sounds which may, or may not, correspond to instruments used in performing this music in a one-to-one mapping. Indeed, the notion of `instruments' often does not make sense and I played one sample in the presentation to try to demonstrate this. Clustering techniques, among other things, are important here.
Furthermore, the simplest conceptual model is to separate the sources, process each of them and them remix them. I want to find ways in which sources be modified without having to explicitly separate them from the mix. This will have ramifications for the computational complexity of the system.
Finally, the blind source separation model seems to be the most popular framework for source separation. This is only one model, and it places very strong, artificial assumptions on the music. So, I am approaching the problem from the perspectives of auditory scene analysis and with consideration to some of the ways in which sound is assembled through the production process."
|
|
Audrey Laplante
audrey.laplante@mail.mcgill.ca
Proposal
Music Information-Seeking Behaviour of Nonspecialists: The Users’ Perspective
|
|
| The aim of my research is to provide deep insights into music information behaviour of nonspecialists. Using the time-line method proposed by Brenda Dervin, I will conduct in-depth interviews with current and potential users of music digital libraries (MDL). Results should inform the design and development of MDL. |
|
Fernando William da Cruz
fwcruz@ucb.br
Proposal
A Brazilian Popular Music Oriented Digital Library for Musical Harmony E-Learning
|
|
| This poster presents a digital library proposal conceived for a virtual community of people interested in acquiring knowledge about Brazilian popular music harmony, particularly in choro. This Brazilian musical style is a complex popular music form based on improvisation, although it contains classical music elements such as the counterpoint. We are proposing two ways of accessing the music virtual library content: a guided navigation mode, in which users interact with a cooperative Web-based learning system; and a free navigation mode, in which users can make their own queries, both through browsers or client applications. This digital library is intended also to be used in a web-based e-learning system. |
|
Maarten Grachten
maarten@iiia.csic.es
Proposal
Respecting Expressivity under Global Musical Tempo Transformations
|
|
| When a skilled musician performs a piece of music at (substantially) different tempos, the expressive characteristics of the performance (timing/dynamics deviations, but also the occurrence of ornamentations etc) are likely to change. These changes are not linear in general. We propose a system to apply the appropriate changes to the expressivity when transforming a performance to a different tempo, based on a set of example transformations. We restrict the music under consideration to monophonic recordings of jazz standards, played by saxophone. |
|
Mitch Parry
parry@cc.gatech.edu
Proposal
Source Separation for Multichannel Music Audio
|
|
| Instrument separation algorithms have been applied to monophonic and stereophonic audio. We propose leveraging the extra information in multichannel audio (i.e. DVD-Audio) for separating instruments. |
|
Nicolas Wack
nwack@iua.upf.es
Proposal
Interactivity in audio browsing, for the user and the researcher
|
|
| We try to describe a general abstract architecture encompassing all fields related to MIR and try to see how they could interact together by the means of software methods. The idea is to give the user the ability to play independently with each one of these categories in a dynamic way, in order for him to improve his research. |
|
Paul Brossier
paul.brossier@elec.qmul.ac.uk
Proposal
Labelling audio objects in real time
|
|
| The aim of this research is to use and investigate MIR algorithms in applications such as live interactive music installations. We focus on the extraction of onset times and of a melodic line bounded by these onsets. Applications range from feature driven effects to music dictation system. |
|
Hugo Solís
hsolis@iua.upf.es
|
|
|
Alfonso Perez
aperez@iua.upf.es
|
|
|
|
