Enhanced blind source separation algorithm for highly correlated mixtures
US8223988B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Jan 29, 2008 |
| Grant date | Jul 17, 2012 |
| Priority date | — |
| Expiry date | May 19, 2031 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG10L2021/02166
- WIPO fieldAudio-visual technology
- WIPO sectorElectrical engineering
Abstract
An enhanced blind source separation technique is provided to improve separation of highly correlated signal mixtures. A beamforming algorithm is used to precondition correlated first and second input signals in order to avoid indeterminacy problems typically associated with blind source separation. The beamforming algorithm may apply spatial filters to the first signal and second signal in order to amplify signals from a first direction while attenuating signals from other directions. Such directionality may serve to amplify a desired speech signal in the first signal and attenuate the desired speech signal from the second signal. Blind source separation is then performed on the beamformer output signals to separate the desired speech signal and the ambient noise and reconstruct an estimate of the desired speech signal. To enhance the operation of the beamformer and/or blind source separation, calibration may be performed at one or more stages.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.