## Blind Source Separation of Real World Signals (1997)

Venue: | Proc. ICNN |

Citations: | 55 - 8 self |

### BibTeX

@INPROCEEDINGS{Lee97blindsource,

author = {Te-won Lee and Anthony J. Bell and Reinhold Orglmeister},

title = {Blind Source Separation of Real World Signals},

booktitle = {Proc. ICNN},

year = {1997},

pages = {2129--2135}

}

### Years of Citing Articles

### OpenURL

### Abstract

We present a method to separate and deconvolve sources which have been recorded in real environments. The use of noncausal FIR filters allows us to deal with nonminimum mixing systems. The learning rules can be derived from different viewpoints such as information maximization, maximum likelihood and negentropy which result in similar rules for the weight update. We transform the learning rule into the frequency domain where the convolution and deconvolution property becomes a multiplication and division operation. In particular, the FIR polynomial algebra techniques as used by Lambert present an efficient tool to solve true phase inverse systems allowing a simple implementation of noncausal filters. The significance of the methods is shown by the successful separation of two voices and separating a voice that has been recorded with loud music in the background. The recognition rate of an automatic speech recognition system is increased after separating the speech signals. 1 Introduct...

### Citations

1365 |
Independent component analysis, a new concept
- Comon
- 1994
(Show Context)
Citation Context ...nications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks [1, 2, 5, 9, 10, 16, 17] and statistical signal processing =-=[3, 7, 11, 15, 19]-=-. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast functions approximated by the Edgeworth expansion of the ... |

514 | A new learning algorithm for blind signal separation
- Amari, Cichocki, et al.
- 1996
(Show Context)
Citation Context ...ir prospective application in speech recognition, telecommunications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks =-=[1, 2, 5, 9, 10, 16, 17]-=- and statistical signal processing [3, 7, 11, 15, 19]. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast func... |

379 | Equivariant adaptive source separation
- Cardoso, Laheld
- 1996
(Show Context)
Citation Context ...kurtosis the sources are sub Gaussian and the term becomes a Hebbian term. It is interesting to note that the direct Bussgang property for blind deconvolution as well as the EASI algorithm by Cardoso =-=[3]-=- lead to a similar learning rule. Efu i+k u i g = Efg(u i )u i+k g \DeltaW / I \Gamma uu T \Gamma g(u)u T (13) 4 FIR Polynomial Filter Design The use of IIR filters is restricted to ARMA (autoregressi... |

298 |
An information maximization approach to blind separation and blind deconvolution
- Bell, Sejnowski
- 1995
(Show Context)
Citation Context ...ir prospective application in speech recognition, telecommunications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks =-=[1, 2, 5, 9, 10, 16, 17]-=- and statistical signal processing [3, 7, 11, 15, 19]. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast func... |

134 | Infomax and maximum likelihood for blind source separation
- Cardoso
- 1997
(Show Context)
Citation Context ...inear mixing of independent variables due to the central limit theorem, the difference between maximizing the distance to the observation or to a Gaussian distribution does not matter in practice. In =-=[4]-=- Cardoso shows that infomax and ML is equivalent because the relation between the KL-distance and the ML differs by the constant Entropy H(x) which is not dependent on W. L = \Gamma Z p(x) log p(x)sp(... |

115 |
A Class of Neural Networks for Independent Component Analysis
- Karhunen, Oja, et al.
- 1997
(Show Context)
Citation Context ...ir prospective application in speech recognition, telecommunications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks =-=[1, 2, 5, 9, 10, 16, 17]-=- and statistical signal processing [3, 7, 11, 15, 19]. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast func... |

96 | A contextsensitive generalization of ICA
- Pearlmutter, Parra
- 1996
(Show Context)
Citation Context |

94 | Blind separation of convolved sources based on information maximisation
- Torkkola
- 1996
(Show Context)
Citation Context ...ve proposed an information theoretic approach to the blind source separation and blind deconvolution problem. This approach has been extended to convolution and time-delays in a feedback architecture =-=[6, 13, 18]-=-. Pearlmutter and Parra have reformulated the ICA in a maximum likelihood (ML) framework [16] where the underlying density is estimated in a context sensitive manner. Although research in blind source... |

79 |
Robust learning algorithm for blind separation of signals
- Cichocki, Unbehauen, et al.
- 1994
(Show Context)
Citation Context |

73 | Multichannel blind deconvolution: FIR matrix algebra and separation of multipath mixtures
- Lambert
- 1996
(Show Context)
Citation Context ...nications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks [1, 2, 5, 9, 10, 16, 17] and statistical signal processing =-=[3, 7, 11, 15, 19]-=-. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast functions approximated by the Edgeworth expansion of the ... |

63 | Blind separation of delayed and convolved sources
- Lee, Bell, et al.
- 1997
(Show Context)
Citation Context ...is given by the gradient ascent of the entropy. @H(y) @w i = @p i (u i ;w i ) @w i p i (u i ; w i ) (11) This allows the separation of sub/super-Gaussian and to some extent Gaussian distributions. In =-=[14]-=- we discuss several alternatives for the estimation of the underlying density. An elegant way of generalizing the learning rule to sub/super-Gaussians is to approximate the estimated pdf in form of th... |

52 |
Multichannel signal separation: methods and analysis
- Yellin, Weinstein
(Show Context)
Citation Context ...nications and medical signal processing. The blind source separation problem has been studied by researchers in the field of neural networks [1, 2, 5, 9, 10, 16, 17] and statistical signal processing =-=[3, 7, 11, 15, 19]-=-. Comon [7] defines the concept of independent component analysis (ICA) which measures the degree of independence among outputs using contrast functions approximated by the Edgeworth expansion of the ... |

38 |
Blind separation of multiple speakers in a multipath environment
- Lambert, Bell
- 1997
(Show Context)
Citation Context .... Although research in blind source separation has been carried out for several years only very few papers have addressed the problem with real acoustic signals recorded in reverberating environments =-=[19, 18, 13, 12]-=-. In this paper, we tackle the problem of separating signals recorded in real environments. The inverting system is approximated by a matrix of finite impulse response (FIR) filters to deconvolve and ... |

36 |
Fast implementations of LMS adaptive filters
- Ferrara
- 1980
(Show Context)
Citation Context ...form of the least mean squared (LMS) adaptive filters. A fast implementation of the LMS adaptive filters in the frequency domain can be archieved by employing the overlap and save block LMS technique =-=[8]-=-. X(z) = FFT[x (k\Gamma1)n \Delta \Delta \Delta x kn\Gamma1 x kn \Delta \Delta \Delta x kn+n\Gamma1 ] (16) For a blocksize of 1024 FFT-points the method is 16 times faster than the conventional LMS me... |

28 |
Multi-dimensional density shaping by sigmoidal networks with application to classification, estimation and forecasting
- Baram, Roth
- 1994
(Show Context)
Citation Context |

20 | Negentropy and kurtosis as projection pursuit indices provide generalised ica algorithms
- Girolami, Fyfe
- 1996
(Show Context)
Citation Context |

17 |
separation of instantaneous mixture of sources via an independent component analysis
- Blind
- 1996
(Show Context)
Citation Context |

9 |
Blind separation of delayed and convolved signals with self-adaptive learning rate
- Cichocki, Amari, et al.
(Show Context)
Citation Context ...ve proposed an information theoretic approach to the blind source separation and blind deconvolution problem. This approach has been extended to convolution and time-delays in a feedback architecture =-=[6, 13, 18]-=-. Pearlmutter and Parra have reformulated the ICA in a maximum likelihood (ML) framework [16] where the underlying density is estimated in a context sensitive manner. Although research in blind source... |

7 |
Blind separation of convolved and delayed sources
- Lee, Bell, et al.
- 1997
(Show Context)
Citation Context ...ve proposed an information theoretic approach to the blind source separation and blind deconvolution problem. This approach has been extended to convolution and time-delays in a feedback architecture =-=[6, 13, 18]-=-. Pearlmutter and Parra have reformulated the ICA in a maximum likelihood (ML) framework [16] where the underlying density is estimated in a context sensitive manner. Although research in blind source... |