#### DMCA

## On the causality problem in time-domain blind source separation and deconvolution algorithms (2005)

### Cached

### Download Links

Venue: | in Proc. IEEE ICASSP |

Citations: | 7 - 6 self |

### Citations

12415 |
Elements of information theory
- Cover, Thomas
(Show Context)
Citation Context ...t channels can be regarded as the most general approach for BSS. To obtain an estimator not only allowing spatial separation but also temporal separation for MCBD, the Kullback-Leibler distance (KLD) =-=[5]-=- between a certain desired joint pdf (essentially representing a hypothesized stochastic source model) and the joint pdf of the actually estimated output signals is used [2]. (ii) Nonwhiteness is expl... |

500 |
Beamforming: A versatile approach to spatial filtering
- Veen, Buckley
- 1988
(Show Context)
Citation Context ...hereas for the source locations in Fig. 1b usually one noncausal demixing filter w12 or w21 is required as will be shown below. Due to the supervised filtering algorithms used in adaptive beamforming =-=[1]-=- the problem of acausality can there be solved by initializing the beamformer FIR filters with a unit impulse at the L/2-th tap (L denotes the filter length). However, unlike for supervised beamformin... |

349 |
Linear Prediction of Speech.
- Markel, Gray
- 1976
(Show Context)
Citation Context ... acausal delays are possible in this case as becomes obvious by simply evaluating (8) for successive iterations. 3.3. Covariance method vs. Correlation method Similarily to linear prediction problems =-=[3]-=- we have to distinguish in actual implementations of the update equation (11) between two methods to estimate the cross-relation matrices defined by N−1 � RyΦ(y)(i) = y H (i, j)Φs,P D(y(i, j)). (17) j... |

120 | Multichannel blind deconvolution and equalization using the natural gradient.
- Amari, Douglas, et al.
- 1997
(Show Context)
Citation Context ...ng the L-th row of ∆W (16) covariance method correlation method assumption of univariate pdfs and neglecting nonstationarity Zhang et al. [8] Joho et al. [11] sample-by-sample algorithms Amari et al. =-=[9]-=- Douglas et al. [12] Douglas et al. [10] Fig. 2. Overview of links between the generic algorithm (11) and existing MCBD algorithms. writing (9) element-wise. When using the first column, i.e., SCC, th... |

52 | Trinicon: A versatile framework for multichannel blind signal processing.
- Buchner, Kellermann
- 2004
(Show Context)
Citation Context ...x1 x2 w11 w21 w12 w22 demixing system Fig. 1. Setups for BSS requiring (a) only causal delays and (b) causal and acausal delays for the demixing system W. be based on a generic framework presented in =-=[2]-=- which contains several different BSS, MCBD, and acoustic dereverberation algorithms. In this paper we will only consider time-domain algorithms as they are not exhibiting typical frequency-domain nar... |

36 | Simultaneous localization of multiple sound sources using blind adaptive MIMO filtering.
- Buchner, Aichner
- 2005
(Show Context)
Citation Context ...iscussed in Sect. 3. It can be seen that it also leads to good separation results for acausal filters. Moreover, the algorithm has been successfully applied to source localization of multiple sources =-=[13]-=-. 5. CONCLUSIONS We examined the problem of acausality in time-domain BSS and MCBD algorithms. Based on a generic framework we showed how different realizations can incorporate acausal demixing filter... |

33 |
Blind source separation for convolutive mixtures: A unified treatment
- Buchner, Aichner, et al.
(Show Context)
Citation Context ...he ensemble average is replaced by time averages over these blocks. For block processing we first need to formulate the convolution of the FIR demixing system of length L in the following matrix form =-=[4]-=-: y(m, j) =x(m, j)W(m), (1) where m denotes the block index, and j = 0, ···,N − 1 is a time-shift index within a block of length N, and x(m, j) = [x1(m, j),...,xP (m, j)], (2) y(m, j) = [y1(m, j),...,... |

33 | Natural gradient multichannel blind deconvolution and speech separation using causal FIR filters
- Douglas, Sawada, et al.
- 2005
(Show Context)
Citation Context ...ethod correlation method assumption of univariate pdfs and neglecting nonstationarity Zhang et al. [8] Joho et al. [11] sample-by-sample algorithms Amari et al. [9] Douglas et al. [12] Douglas et al. =-=[10]-=- Fig. 2. Overview of links between the generic algorithm (11) and existing MCBD algorithms. writing (9) element-wise. When using the first column, i.e., SCC, then the matrix multiplication (9) results... |

21 | Multichannel blind deconvolution of nonminimum-phase systems using filter decomposition
- Zhang, Cichocki, et al.
- 2004
(Show Context)
Citation Context ...re ∆T corresponds to the gradient descent in the tangent search space which is projected back to the original manifold of the Euclidean space by the multiplication with W (for more details see, e.g., =-=[6]-=-). This leads to the following generic TRINICON-based update rule which can be written equivalently to [2] as: W(m) = W(m − 1) − µ∆W(m), (10) ∆W(m) = 2 ∞� β(i, m)W(i) N i=0 N−1 � � � H · y (i, j)Φs,P ... |

11 | Frequency domain realization of a multichannel blind deconvolution algorithm based on the natural gradient
- Joho, Schnitter
- 2003
(Show Context)
Citation Context ... Enforcing the Sylvester constraint SCR by using the L-th row of ∆W (16) covariance method correlation method assumption of univariate pdfs and neglecting nonstationarity Zhang et al. [8] Joho et al. =-=[11]-=- sample-by-sample algorithms Amari et al. [9] Douglas et al. [12] Douglas et al. [10] Fig. 2. Overview of links between the generic algorithm (11) and existing MCBD algorithms. writing (9) element-wis... |

10 | Real-time convolutive blind source separation based on a broadband approach
- Aichner, Buchner, et al.
- 2004
(Show Context)
Citation Context ...sing a Sylvester constraint SC. This constraint can be enforced by selecting the L filter taps in the first column of ∆W as a reference and generate the Sylvester structure in the form of (8) from it =-=[4, 7]-=-. This implementation is denoted by SCC. Another option for enforcing the Sylvester constraint is to select the L-th row of ∆W as a reference (denoted by SCR) and generate the Sylvester structure in t... |

3 |
A causal frequency-domain implementation of a natural gradient multichannel blind deconvolution and source separation algorithm
- Douglas, Sawada, et al.
- 2004
(Show Context)
Citation Context ...∆W (16) covariance method correlation method assumption of univariate pdfs and neglecting nonstationarity Zhang et al. [8] Joho et al. [11] sample-by-sample algorithms Amari et al. [9] Douglas et al. =-=[12]-=- Douglas et al. [10] Fig. 2. Overview of links between the generic algorithm (11) and existing MCBD algorithms. writing (9) element-wise. When using the first column, i.e., SCC, then the matrix multip... |

1 |
et al., “Geometrical structures of FIR manifold and their application to multichannel blind deconvolution
- Zhang
- 1999
(Show Context)
Citation Context ...rrelation method Enforcing the Sylvester constraint SCR by using the L-th row of ∆W (16) covariance method correlation method assumption of univariate pdfs and neglecting nonstationarity Zhang et al. =-=[8]-=- Joho et al. [11] sample-by-sample algorithms Amari et al. [9] Douglas et al. [12] Douglas et al. [10] Fig. 2. Overview of links between the generic algorithm (11) and existing MCBD algorithms. writin... |