Source Separation of Baseband Signals in Post-Nonlinear Mixtures

Abstract : Usually, source separation in Post-Nonlinear (PNL) models is achieved via one-stage methods, i.e. the two parts (linear and nonlinear) of a PNL model are dealt with at the same time. However, recent works have shown that the development of two-stage techniques may simplify the problem. Indeed, if the nonlinear stage can be compensated separately, then, in a second moment, one can make use of the well-established source separation algorithms for the linear case. Motivated by that, we propose in this work a novel two-stage PNL method relying on the assumption that the sources are bandlimited signals. In the development of our method, special care is taken in order to make it as robust as possible to noise. Simulation results attest the effectiveness of the proposal.
Complete list of metadatas

Cited literature [9 references]  Display  Hide  Download
Contributor : Leonardo Tomazeli Duarte <>
Submitted on : Saturday, September 12, 2009 - 1:02:38 PM
Last modification on : Monday, July 8, 2019 - 3:10:15 PM
Long-term archiving on : Tuesday, June 15, 2010 - 11:35:05 PM


Files produced by the author(s)


  • HAL Id : hal-00416161, version 1


Leonardo Tomazeli Duarte, Christian Jutten, Bertrand Rivet, Ricardo Suyama, Romis Attux, et al.. Source Separation of Baseband Signals in Post-Nonlinear Mixtures. IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2009), Sep 2009, Grenoble, France. CD. ⟨hal-00416161⟩



Record views


Files downloads