Projects

Hierarchical Segmentation & Semantic Markup of Musical Signals (EPSRC funded)
Automatic segmentation and classification of music signals. These technologies are required for media asset management for on-line music services, media archiving services and for consumer music devices. We have developed new signal processing algorithms that perform meaningful segmentations of musical signals for these applications. In our framework, the higher-order patterns in a piece of music can be extracted at many hierarchical levels such as beat, bar, phrase and movement.

Information Dynamics of Music (EPSRC funded)
The aim of this project is to develop a novel data-driven methodology for automatically analysing the dynamic structure of music. The project will focus on the idea of expectation and surprise in music, and we will investigate how these can be quantified in terms of measures derived from information theory, related to the hypothesis of redundancy reduction in perception. We anticipate that this project will produce important new insights into the structure of music and other time-based sionals. and could gave the wav for significant future advances in music analvsis and processing.

OMRAS2: A Distributed Research Environment for Music Informatics and Computational Musicology (EPSRC funded)
Online Music Recognition and Searching. We are developing algorithms for exploring large corpuses of music by characteristics of the sound of the music directly, rather than by linguistic description. They will also help music researchers investigate interesting aspects of music, such as what variations of that riff in Purple Haze did Jimi Hendrix play and how did they differ, and how did different pianists interpret Bach's Goldberg Variations. In the end we will have an intuitively navigable space in which to explore vast quanities of music based on the qualities of the sound.

Modelling Melodic Memory and the Perception of Musical Similarity (EPSRC funded)
One of the ways music listeners often find their music is by searching for songs or pieces that sound like one another. The aim of our project is to understand the different ways in which people view different pieces of music as similar and to build a computer program which is capable of making the same judgements. That programme can then be used to help people find the music they want in shops, libraries and so on. Perhaps the most important outcome of the project, which underlies all the above, is a better understanding of one aspect of how the human mind works. Thus, the project aims to help us understand ourselves.

Efficient Data Structures and Algorithms for Navigating and Managing Very Large Music Collections (funded by Yahoo! Research Alliance).
This research follows up on Online Music Recognition and Searching (OMRAS2). We estimate that the algorithm, developed by Michael Casey and Malcolm Slaney, saves a factor of 1,000,000,000 on music similarity comparisons in their database of two million songs.

Mutators (funded by the AHRC, The Arts Council, and the Emerald Fund)
This project follows on seminal computer art work that William Latham did in the 1970's and 19080's. William is working with us, updating those ideas with emerging three-dimensional modelling ideas to produce algorithms for geometric design and computer visualisation. Applications of the Designer work include Computer Art the automatic generation of 3D content for the Games, Special Effects (Film+TV), and CAD-CAM industries; applications for the visualisation include visualisation of complex numerical sequences including genomics, proteomics, financial data streams

New Millennium New Media (EU Funded)
We are developing tools, techniques, and strategies for creating and for engaging with non-linear narratives. Partners on the project include: BT, BBC, Sony, Finnish television. Goldsmiths is principally responsible for the reasoning software, software the reasons about how to configure fragments of a story into a coherent narrative, and the high-level production tools.

Culture Mining (AHRC-funded)
A project with the Tate, in which we are turning the Tate's streaming media collection into a research archive by creating an ontology for contemporary art and producing algorithms for semi-automatically fragmenting parts of the collection and tagging the fragments with meta-data associated with the ontology.

Modelling Adult Stem Cells using Multi-Agent Systems (Wellcome trust funded)
In collaboration with several European, Australian and American partners, agent-based modelling to understand stem cell behaviour. As part of this project we are working with the artist Jane Prophet to find ways of presenting different visualizations of this behaviour to different interested communities. This project also includes the stem cell researcher Neil Theise, the ALife programmer Rob Saunders and the curator Peter Ride.

New partner funding scheme with industrial and cultural partners (funded by Arts and Business).
This umbrella funding brings together performing artists, visual artists and companies to generate research ideas. We are the convenors and the other participants include, or have included, BT, The London Philharmonia, Dance East, Cambridge University, and FACT.

Hexagram
A multi-disciplinary research institute based in Montreal supports two projects, through Canadian government funding on artistic and social uses of clothes with embedded computational devices. The two projects are called Intelligent textiles and Wearable Absence.

Intimate Technologies (EPSRC funded)
Touching Textiles to improve the quality of human-computer Haptic (touch and feel) interaction in Material Culture. This work has been funded by the EPSRC and a joint project with Mandayam A. Srinivasan, director of the Touch Lab at MIT.