Experimental comparison of signal subspace based noise reduction methods

Hansen, Poul Henrik Kyvsgaard; Hansen, P. N.; Hansen, Søren; Sorensen, John A.

doi:10.1109/icassp.1999.758072

Cited by 11 publications

(12 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To see that the prewhitened noise is indeed white, we note that is the covariance matrix for white noise in the subspace . As shown in [12] and [13], different optimality criteria lead to different formulas for the reconstructed signal, and a common feature is that the filtering is achieved by multiplying the singular values with appropriate factors. Hence, to compute the filtered matrix via the QSVD, we first modify the singular values of the matrix quotient and then right-multiply with .…”

Section: A Full-rank Prewhitenermentioning

confidence: 99%

“…Hence, to compute the filtered matrix via the QSVD, we first modify the singular values of the matrix quotient and then right-multiply with . Inserting the QSVD, it is easy to see that the complete process can be written as where denotes a diagonal filter matrix [12], [13]. It follows immediately that the covariance matrix for the filtered signal is given by…”

Section: A Full-rank Prewhitenermentioning

confidence: 99%

“…To incorporate filtering, we must filter or truncate this submatrix. Hence, the QULVbased reconstruction takes the form where is the filter matrix [12].…”

Section: A Ulv and Qulv Decompositionsmentioning

confidence: 99%

See 2 more Smart Citations

Prewhitening for rank-deficient noise in subspace methods for noise reduction

Hansen

Jensen

2005

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

Abstract-A fundamental issue in connection with subspace methods for noise reduction is that the covariance matrix for the noise is required to have full rank in order for the prewhitening step to be defined. However, there are important cases where this requirement is not fulfilled, e.g., when the noise has narrowband characteristics or in the case of tonal noise. We extend the concept of prewhitening to include the case when the noise covariance matrix is rank deficient, using a weighted pseudoinverse and the quotient singular value decomposition, and we show how to formulate a general rank-reduction algorithm that works also for rank-deficient noise. We also demonstrate how to formulate this algorithm by means of a quotient ULV decomposition, which allows for faster computation and updating. Finally, we apply our algorithm to a problem involving a speech signal contaminated by narrowband noise.

show abstract

Section: A Full-rank Prewhitenermentioning

confidence: 99%

Section: A Full-rank Prewhitenermentioning

confidence: 99%

See 1 more Smart Citation

Prewhitening for rank-deficient noise in subspace methods for noise reduction

Hansen

Jensen

2005

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…For detailed derivations of equations (2) and (3), please refer to [2,4,8]. Note that only q singular values are used, i.e.…”

Section: Pink and Babble Noises In Imfsmentioning

confidence: 99%

Speech enhancement of color noise using empirical mode decomposition

Koh

Rodriguez-Marek

2012

2012 Conference Record of the Forty Sixth Asilomar Conference on Signals, Systems and Computers (ASILOMAR)

View full text Add to dashboard Cite

This paper introduces a new speech enhancement algorithm, which is efficient for color noises such as pink and babble noises. The algorithm applies the minimum variance estimator (MVE) into each intrinsic mode function (IMF) obtained by the empirical mode decomposition (EMD). Typical signal subspace approaches to speech enhancement assume the corrupting noise is white, thereby making a pre-whitening step mandatory when faced with various color noises. The algorithm presented in this paper is applicable into color noises, such as pink and babble noises, without the need to apply the whitening process. This is a result of the noises' variances in each IMF signal showing smaller fluctuations than without the application of the EMD. Experimental data demonstrates the effectiveness of the algorithm to combat color noises.

show abstract

“…Singular value decompositions of two-dimensional data matrices are also employed in noise reduction for speech enhancement (see [16] and references therein) and the design of a two-dimensional filter ( [15] …”

Section: Decomposition Of Three-dimensional Image Setsmentioning

confidence: 99%

State Space Realization of a Three-dimensional Image Set with Application to Noise Reduction of Fluorescent Microscopy Images of Cells

Ober

Lai

Lin

2005

Multidim Syst Sign Process

View full text Add to dashboard Cite

Abstract. A method is presented to calculate state space realizations of a three-dimensional image set. It is based on interpreting the image set as the impulse response of a 3D separable system. As an application it is shown how this method, combined with approximation steps, including balanced model reduction, can be used to suppress noise in three-dimensional image sets. The approach was motivated by a practical problem in the analysis of three-dimensional fluorescent microscopy image data of fluorescently labelled cells. The method is illustrated by an analysis of simulated data and experimental data. The proposed approach can also be applied to a two-dimensional image in a straightforward way.

show abstract

Experimental comparison of signal subspace based noise reduction methods

Cited by 11 publications

References 5 publications

Prewhitening for rank-deficient noise in subspace methods for noise reduction

Prewhitening for rank-deficient noise in subspace methods for noise reduction

Speech enhancement of color noise using empirical mode decomposition

State Space Realization of a Three-dimensional Image Set with Application to Noise Reduction of Fluorescent Microscopy Images of Cells

Contact Info

Product

Resources

About