Music Technology Forum Page
Researchers, postgraduate students and composers (mainly working at the University of York) constitute the Music Technology Forum.
It is a very informal group that meets usually every month and organises, while having coffee!, events about Music Technology.
Events:
17 November 2004 (1.15 – 17.15), G/045 Goodricke College, The University of York
The mapping of gestures, sounds, visuals and structures:
research and applications
An afternoon workshop with researchers from the electronics and the Music Department.
Speakers:
Andy Hunt, Media Engineering Group, University of York
Matt Neighbour, Media Engineering Group, University of York
Matthew Paradis, Media Engineering Group and Music Department, University of York
Sandra Pauletto, Media Engineering Group, University of York
Roger Marsh, composer, Music Department, University of York
15 April 2004 (9.30 – 16.30)
Time-Frequency Analysis for
Audio: theory – applications - methods
A one day workshop in the Department of Electronics
Speakers:
Jez Wells, Media Engineering Group, University of York
Mark Every, Media Engineering Group, University of York
Dr John Szymanski, Media
Engineering Group, University of York
Fritz Menzer, École
Polytech. Féd. De Lausanne, Switzerland
Trevor Wishart,
Electroacoustic composer
Some photos from the
workshop:
7 June 2004
Research seminar in the Electronics Department of the
University of York
The speaker: James Mooney, The university of Sheffield sound studios.
The title: The M2 Diffusion System
3 May 2004
The speaker: Thomas Blumensath, Queen Mary University, London
Title: Unsupervised learning of shift-invariant
representations: A new linear model for music analysis
Abstract:
Music is a highly structured signal. Most algorithms for music analysis use
knowledge of these structures such as harmonic relationships. The structures
are, however, only known approximately in many instances and a complete
specification is often not feasible.
Machine learning techniques can discover the underlying structure of data in an
unsupervised framework. These approaches rely on the specification of a model
generating the data, which is often less restrictive then other methods.
Working in a Bayesian statistical framework offers a formal way of learning
model parameters and inferring unknown causes of the data.
In this talk a model will be proposed in which musical signals are mixtures of
individual instrument sounds observed with noise. Two assumptions will be used
to disentangle these mixtures. It can be assumed that the sounds of each
instrument or note will be played independently from other notes in the mixture
and that only a few notes will be played at each instance in time. These are
the underlying assumptions of the standard sparse coding model. As musical
signals are not restricted to certain time locations this model has to be
implemented in a shift-invariant framework in which notes can occur at
arbitrary time location. It can be shown that such a framework can be used for
a range of applications such as polyphonic music transcription, blind source
separation and de-noising.
9 February 2004
Research seminar in the Electronics Department of the University of York
The Speaker: Ambrose Field, Music Technology, Department of Music.
The Title: One hell of a place to lose a cow.
Abstract:
The musical work "One hell of a place to lose a cow" by Ambrose
Field was the winner of the International Studio Music PRIX at the Boruges
Festival in France, 2003. Described by Steve Benner of Sonic Arts network
diffusion magazine as "an incredible hard-rock sonic massage", the
piece takes sound design to new levels. This seminar looks at the creation of
the work, the role of technology in making some of the sounds, and the final presentation
of the piece in 5.1 surround sound.