|
| EPSRC Reference: |
EP/C005554/1 |
| Title: |
NETWORK: Blind Source Separation and Independent Component Analysis Network |
| Principal Investigator: |
Professor M Plumbley |
| Other Investigators: |
|
| Researcher Co-investigator: |
|
| Project Partner: |
|
| Department: |
Sch of Electronic Eng & Computer Science |
| Organisation: |
Queen Mary, University of London |
| Scheme: |
Standard Research |
| Starts: |
01 April 2005 |
Ends: |
30 September 2008 |
Value (£): |
62,058
|
| EPSRC Research Topic Classifications: |
| Digital Signal Processing |
Multimedia |
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary |
The aim of this network is to improve communication within the UK in the interdisciplinary area of blind source separation and independent component analysis (BSS/ICA).
In many signal and data analysis situations, observed data are known to be some mixture of underlying sources. The mixing process may be linear or nonlinear, and while the structure of the mixing process may be known, the mixture parameters (in the linear case, the mixing matrix) will be unknown. This problem is therefore often known as blind source separation (BSS). For a simple audio example, we may have two speakers in a room, with two microphones each receiving a different mixture of the two speakers. The task is then to recover the original (unmixed) speakers from the two mixtures received at the microphones.
The BSS problem can be tackled using Independent Component Analysis (ICA) and related techniques. ICA assumes the underlying sources are statistically independent from each other; related techniques may assume few sources are non-zero at any time (sparse coding) or the sources and/or mixtures must be non-negative (non-negative factor analysis). Applications of these techniques have been demonstrated in the analysis of EEG signals, MRI spectra, computer vision, gene microarray data, fMRI images, text document collections, satellite images and economic data, and it has been applied to watermarking for information hiding, speech enhancement in noisy or echoic environments, and image coding.
The activities of the Network will include: workshops and meetings, visits to conferences and other laboratories (particularly for younger researchers), and electronic communication such as email and a web site. An important Network activity will be the production of a "Research Roadmap" to help future research planning and inform funding bodies on the state of the art and future opportunities. Dissemination of activities and information relevant to the Network will be an important aspect of its work, leading to increased opportunities for exploitation and collaboration with UK industry and international researchers. Other relevant groups and researchers will be encouraged to join the Network during its operation. The proposed Network is expected to lead to increased collaborative activity in the field.
|
| Final Report Summary |
The aim of this network is to improve communication within the UK in the interdisciplinary area of blind source separation and independent component analysis (BSS/ICA), and improve the international visibility and impact of UK reserach in this area.
In many signal and data analysis situations, observed data are known to be some mixture of underlying sources. The mixing process may be linear or nonlinear, and while the structure of the mixing process may be known, the mixture parameters (in the linear case, the mixing matrix) will be unknown. This problem is therefore often known as blind source separation (BSS). For a simple audio example, we may have two speakers in a room, with two microphones each receiving a different mixture of the two speakers. The task is then to recover the original (unmixed) speakers from the two mixtures received at the microphones.
The BSS problem can be tackled using Independent Component Analysis (ICA) and related techniques. ICA assumes the underlying sources are statistically independent from each other; related techniques may assume few sources are non-zero at any time (sparse coding) or the sources and/or mixtures must be non-negative (non-negative factor analysis). Applications of these techniques have been demonstrated in the analysis of EEG signals, MRI spectra, computer vision, gene microarray data, fMRI images, text document collections, satellite images and economic data, and it has been applied to watermarking for information hiding, speech enhancement in noisy or echoic environments, and image coding.
The activities of the Network will include: workshops and meetings, visits to conferences and other laboratories (particularly for younger researchers), and electronic communication such as email and a web site. Dissemination of activities and information relevant to the Network will be an important aspect of its work, leading to increased opportunities for exploitation and collaboration with UK industry and international researchers. Other relevant groups and researchers will be encouraged to join the Network during its operation. The proposed Network is expected to lead to increased collaborative activity in the field.
|
| Further Information: |
|
| Organisation Website: |
http://www.qmul.ac.uk |
|
|