Presentation | 2019-04-19 Blind Source Separation in Nonlinear Mixture: Separation and a Multi-Subspace Representation Lu Wang, Tomoaki Ohtsuki, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | The process deals with blind source separation in the nonlinear domain is to estimate the original signals or mixture functions from the degraded signals, without any prior information about the mixing functions. The fundamental problem is to recover the original sources by estimating an approximation function under such assumptions so as to estimate the inverse of mixing functions. However, in practice, the approximation function is derived from some estimation algorithm with a finite sample size that even larger estimation error appears with improper model construction. In this paper, we work on the convergence and asymptotic analysis of the separation approach, where the nonlinearity of the mixture function is extracted by the flexible approximation and the nonlinear problem is solved linearly in the feature space. The analysis stems from the performance of a mismatched estimator that accesses the finite sample size. By providing a closed-form expression of the mean squared error (MSE), we can present a novel algebraic formalization as well as derive an upper bound on the estimation error. The simulation results show that if the nonlinearity of mixing functions can be extracted by the flexible approximation, the consistency of numerical MSE and analytical MSE can be achieved as the sample size tends to be infinity. This implies that the algorithm is feasible to separate the distortion of the nonlinear mixture. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Nonlinear blind source separationmulti-subspace representationnonlinear mixturekernel mappingtime-frequency representation |
Paper # | RCS2019-16 |
Date of Issue | 2019-04-11 (RCS) |
Conference Information | |
Committee | RCS |
---|---|
Conference Date | 2019/4/18(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Noboribetsu Grand Hotel |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Railroad Communications, Inter-Vehicle Communications, Road to Vehicle Communications, Radio Access Technologies, Wireless Communications, etc. |
Chair | Tomoaki Otsuki(Keio Univ.) |
Vice Chair | Eisuke Fukuda(Fujitsu Labs.) / Satoshi Suyama(NTT Docomo) / Fumiaki Maehara(Waseda Univ.) |
Secretary | Eisuke Fukuda(Hokkaido Univ.) / Satoshi Suyama(NTT) / Fumiaki Maehara |
Assistant | Kazushi Muraoka(NTT DocomoO) / Shinsuke Ibi(Osaka Univ.) / Hiroshi Nishimoto(Mitsubishi Electric) / Koichi Adachi(Univ. of Electro-Comm.) / Osamu Nakamura(Sharp) |
Paper Information | |
Registration To | Technical Committee on Radio Communication Systems |
---|---|
Language | ENG |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Blind Source Separation in Nonlinear Mixture: Separation and a Multi-Subspace Representation |
Sub Title (in English) | |
Keyword(1) | Nonlinear blind source separationmulti-subspace representationnonlinear mixturekernel mappingtime-frequency representation |
1st Author's Name | Lu Wang |
1st Author's Affiliation | Keio Univeristy(Keio Univ.) |
2nd Author's Name | Tomoaki Ohtsuki |
2nd Author's Affiliation | Keio Univeristy(Keio Univ.) |
Date | 2019-04-19 |
Paper # | RCS2019-16 |
Volume (vol) | vol.119 |
Number (no) | RCS-8 |
Page | pp.pp.73-78(RCS), |
#Pages | 6 |
Date of Issue | 2019-04-11 (RCS) |