Funded by

ICArn 1-day Workshop on Applied BSS/ICA

ICA Research Network (ICArn) Workshop on Applied Blind Source Separation and Independent Component Analysis

University of Southampton
23 September 2005

Contents: Introduction | Registration | Travel Funding | Programme | Venue | Hotels | Posters | Abstracts

Introduction

Following the success of the Launch Day, the ICA Research Network (ICArn) will be holding a 1-day Workshop on Applied BSS/ICA at the University of Southampton on 23 September 2005.

The workshop will focus on the applications of blind source separation (BSS) techniques such as independent component analysis (ICA) to a range of application areas, including (but not restricted to):

• Biomedicine
• Audio/speech/music
• Commerce and finance
• Psychology

Registration to this workshop will be free (but you must register). There will be a number of invited speakers who will highlight techniques used in these application fields. There will be ample opportunity for networking throughout the day.

Invited talks include:

• Dr Ricardo Vigário (Helsinki University of Technology) on
  ICA in Neuroinformatics
• Dr Emmanuel Vincent (Queen Mary University of London) on
  Blind Audio Source Separation - A critical review
• Dr Daniel Osorio (University of Sussex) on
  Control of coloration by cuttlefish

Deadlines

• 1 Sept 2005: Abstract submission deadline for posters
• 8 Sept 2005: Notification of poster acceptance
• 16 Sept 2005: Registration (free)
• 23 Sept 2005: Workshop

Call for Proposals: Posters

You are invited to submit a one (A4) page abstract about the application of BSS/ICA to your field of interest. We strongly encourage each partner in the network to submit at least one abstract. Accepted abstract authors will be invited to produce a poster (A0) which will be displayed throughout the event.

The abstracts will be collated into a digest and distributed on the day, and authors will be encouraged to submit an electronic versions of posters (e.g. in PDF format) to allow the posters to be viewed after the event.

Abstract Submission

Please submit your poster abstract (maximum 1 page of A4) in an email to Mark Plumbley at mark.plumbley@elec.qmul.ac.uk giving the Authors, Title and Abstract of your poster.

Abstract submission deadline: 1 September 2005.

You will be notified of poster acceptance by 8 September 2005.

Poster Instructions

Each poster has to fit on a poster board that is 3 feet (91.4 cm) wide and 6 feet (182.9 cm) tall. However, posters should not reach down to the floor as this makes them hard to read. Posters should therefore be no more than 85 cm (33.5 in) wide and no more than 119 cm (46.9 in) tall (i.e., no larger than A0 portrait or A1 landscape).

IMPORTANT: Posters wider than the stated dimensions will not fit on the poster boards. A0 landscape is TOO WIDE.

Registration

Thanks to financial support from EPSRC, the event is free, but delegates must register (to plan rooms and catering). Places are strictly limited, and will be allocated on a first-come, first-served basis.

To register, please the following information in an email to theresa.willis@elec.qmul.ac.uk:

1. The event ("ICArn Worshop on Applied BSS/ICA")
2. Your Name (Title, First name, Last name)
3. University/company affiliation (for badge)
4. Postal address (including postcode)
5. Your Email address
6. Any dietary requirements (vegetarian, etc.)

Registration deadline: 16 September 2005

Travel Funding for Young Researchers

Some travel funding will be available for PhD students and other young researchers based in the UK. To apply, see the Travel Support page.

Provisional Programme

The programme will include time for coffee, lunch and tea breaks to view posters and meet other researchers in the network.

Key: SC - Staff Club; LT - Shackleton Building Lecture Theatre

Venue

The presentations will be given in the Shackleton Building Lecture Theatre, on the Highfield Campus of the University of Southampton.

• Highfield Campus (including travel)
• University of Southampton: Campus Locations

Hotels

When booking hotels, please advise them that you are visiting the University: many of the hotels offer special rates.

Due to the Southampton Boat Show, you may need to contact a few hotels, since some hotels near the City Centre may already be full.

Hotels near to Venue:

• Premier Travel Inn, Southampton Airport
• Elizabeth House Hotel, The Avenue
• Hilton Southampton, Chilworth
• Highfield House Hotel (Self catering)

City Centre hotels:

• Star Hotel
• Southampton Park
• Premier Travel Inn, Southampton City Centre
• Travelodge, Southampton
• Novotel, Southampton
• Ibis Southampton Centre
• Holiday Inn Southampton
• Dolphin Hotel

Further out:

• Holiday Inn, Eastleigh
• Holiday Inn Express, Nursling
• Premier Travel Inn, Southampton North, Nursling
• Premier Travel Inn, Southampton West, Ower
• Travelodge, Eastleigh
• Travelodge, Stoney Cross
• Bartley Lodge, Cadnam

Posters

Bayesian approaches to time-frequency based blind audio source separation
Cédric Févotte and Simon J. Godsill
Cambridge University Engineering Dept

Using Video to Solve the Cocktail Party Problem
A. Aubrey, Y. Hicks, J. Chambers, S. Sanei and D. Marshall
Cardiff University

Recent research at Edinburgh in BSS and ICA
James R. Hopgood and Bernie Mulgrew
University of Edinburgh

A maximum likelihood algorithm for slow feature analysis
Richard Turner and Maneesh Sahani
Gatsby Institute, UCL

Latent Process Decomposition of Microarray Data
Simon Rogers and Mark Girolami
University of Glasgow

Adaptive Image fusion using ICA bases
Nikolaos Mitianoudis and Tania Stathaki
Imperial College London

Speech Modelling with Application to Convolutive Blind Source Separation
Kostas Kokkinakis and Asoke K. Nandi
University of Liverpool

Blind OFDM Receivers based on Independent Component Analysis for Multiple-Input Multiple-Output Systems
Luciano Sarperi, Xu Zhu and Asoke K. Nandi
University of Liverpool

Towards Agent-Based Independent Component Analysis
Roman Belavkin
Middlesex University

Tensorial extensions to ICA for multi-subject/session FMRI data
Christian Beckman
University of Oxford

ICA in a Computer Game
Colin Fyfe
University of Paisley

BSS on real data: tracking and implementation
Paul Baxter
Qinetiq

Speech and music applications of independent component analysis
Maria Jafari et al
Queen Mary University of London

Temporal ICA for automatic artefact removal from EEG
N. Nicolaou and S. J. Nasuto
University of Reading

Tracking Epileptiform Activity in the Multichannel Ictal EEG using Spatially Constrained Independent Component Analysis
Christian Hesse, Christopher James
University of Southampton

Independent component analysis, a new framework for speech processing of cochlear implant?
Guoping LI, Mark E Lutman
University of Southampton

Abstracts

ICA in Neuroinformatics
Ricardo Vigário
Helsinki University of Technology

Biomedical applications are arguably the most successful practical uses of Independent Component Analysis. During the presentation, I will pass in review early research carried out in the Neural Networks Research Centre. We will cover topics such as artifact removal in EEG, MEG and fMRI; the analysis of even related activity; the introduction of prior information to ICA, in search for more physiologically plausible components; the study of cortico-muscle interaction. The assessment of the reliability of the ICA decomposition will also be addressed, supported by a case study in fMRI.

Blind Audio Source Separation - A critical review
Emmanuel Vincent
Centre for Digital Music, Queen Mary University of London

Blind Audio Source Separation (BASS) aims to recover the waveform of each source within an audio signal containing several audio sources active simultaneously. This has many direct applications, including denoising for auditory protheses, post-production of raw musical recordings and rendering of stereo data on multichannel devices.

Early work on BASS includes Independent Component Analysis (ICA), which uses a generic statistical model to separate multi-channel signals with low reverberation, and Computational Auditory Scene Analysis (CASA), which exploits specific properties of audio sources such as harmonicity and timbre to separate single-channel signals. More recently, statistical models using time-frequency masking have been proposed for the separation of stereo signals. In addition, advanced statistical source models, such as factorial hidden Markov models and sparse nonnegative factorization of the power spectrogram, have also been applied to single- and multi-channel BASS.

This talk will provide a critical review of these established and recent methods as applied to real world BASS problems. The different approaches will be compared within a unified statistical framework and the special features of audio data will be pointed out. Finally a list of open problems will be proposed for further research and promising techniques such as probabilistic waveform modeling will be mentioned.

Control of coloration by cuttlefish
Daniel Osorio & John Anderson
School Of Life Sciences, University of Sussex

Ordinary behaviour requires the action of many muscles controlled by the nervous system. One can conceive behaviours as being produced by a number of discrete neural centres, whose outputs can in some cases be mixed. In visual communication well-known examples are found in the expression of emotion by man and animals, where a range of expressions are produced by mixing a few basic states - fear, aggression, pleasure and disgust. We are using ICA to investigate the related problem of how the common cuttlefish co-ordinates the expression of some 50 separate visual and postural features to produce a highly flexible system for camouflage and communication. In principle ICA is well-suited to this problem as the patterns are too complex to analyse visually but combine approximately linearly to determine the visual appearance of the animals body.