Bias compensation methods for minimum statistics noise power spectral density estimation

Martin, Rainer

doi:10.1016/j.sigpro.2005.07.037

Cited by 66 publications

(53 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To correct for such a bias in the estimated noise PSD as a result of consistent over or underestimates of , we introduce a signal subspace dimension dependent bias compensation factor and compute as (12) The argumentation that we use to define the bias compensation factor is similar to the one introduced in [22]. The use of this bias compensation factor is based on the fact that (13) is proportional to .…”

Section: B Bias Compensation Ofmentioning

confidence: 99%

Noise Tracking Using DFT Domain Subspace Decompositions

Hendriks

Jensen

Heusdens

2008

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-All discrete Fourier transform (DFT) domain-based speech enhancement gain functions rely on knowledge of the noise power spectral density (PSD). Since the noise PSD is unknown in advance, estimation from the noisy speech signal is necessary. An overestimation of the noise PSD will lead to a loss in speech quality, while an underestimation will lead to an unnecessary high level of residual noise. We present a novel approach for noise tracking, which updates the noise PSD for each DFT coefficient in the presence of both speech and noise. This method is based on the eigenvalue decomposition of correlation matrices that are constructed from time series of noisy DFT coefficients. The presented method is very well capable of tracking gradually changing noise types. In comparison to state-of-the-art noise tracking algorithms the proposed method reduces the estimation error between the estimated and the true noise PSD. In combination with an enhancement system the proposed method improves the segmental SNR with several decibels for gradually changing noise types. Listening experiments show that the proposed system is preferred over the state-of-the-art noise tracking algorithm.Index Terms-Discrete Fourier transform (DFT) domain subspace decompositions, noise tracking, speech enhancement.

show abstract

Section: B Bias Compensation Ofmentioning

confidence: 99%

Noise Tracking Using DFT Domain Subspace Decompositions

Hendriks

Jensen

Heusdens

2008

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…The noise statistics and the noise vectors were adaptively updated during the periods where the speech signal is absent. The noise statistics can be further improved by combining other relevant noise estimation techniques incorporating the scheme such as the minimum statistics or noise flooring, 11,12 For all our experiments, we used the Gaussian kernel function because it has been widely used in the kernel related VAD approaches. 6 It is defined as kðy i ; y j Þ ¼ exp½Àðjjy i À y j jj 2 =2r 2 Þ where r is the kernel parameter to control the width of the Gaussian kernel.…”

Section: Implementation and Resultsmentioning

confidence: 99%

Enhanced voice activity detection in kernel subspace domain

Kim

Shin

Chang

2013

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

This paper proposes a voice activity detection (VAD) method in a kernel subspace domain to improve the performance of the kernelbased VAD. A linear transform matrix that can simultaneously diagonalize the two covariance matrices using kernel principal component analysis is presented to generate the kernel subspace. The likelihood ratio test based on Gaussian distributions is applied for the VAD in the kernel subspace. Experimental results show that the proposed VAD algorithm outperforms the conventional approaches under various noise conditions.

show abstract

“…The noise statistics can be further improved by combining other relevant noise estimation techniques incorporating the scheme of noise flooring. 10 For all our experiments, we used the Gaussian kernel function because it has been widely used due to its superior capability under the nonlinear classification problem in the applications of SVM-based VAD and kernel-based speech enhancement. 5,6 It is defined as…”

Section: Implementation and Resultsmentioning

confidence: 99%

“…8 It is necessary to compute the log likelihood of the training samples {/(y 1 ),…,/(y N )} under the approximated GD in Eq. (10). It was shown 8 that the log-likelihood function is given by…”

Section: Gaussian Density In Kernel Spacementioning

confidence: 99%