Using information theoretic distance measures for solving the permutation problem of blind source separation of speech signals

Hoffmann, Eugen; Kolossa, Dorothea; Köhler, Bert-Uwe; Orglmeister, Reinhold

doi:10.1186/1687-4722-2012-14

Cited by 11 publications

(14 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, STFT coefficients can be modelled in accordance with the Chi [20] and Rayleigh [17] distribution functions. Moreover, the logarithm of the STFT speech magnitude is modelled by a generalized Gaussian distribution in [17]. However, a single pdf class cannot model the distribution of STFT coefficients with high accuracy across all frequencies.…”

Section: Measuring the Similarity Of The Speech Spectrummentioning

confidence: 99%

See 1 more Smart Citation

A Contrast Function Based on Generalized Divergences for Solving the Permutation Problem in Convolved Speech Mixtures

Sarmiento

Durán-Díaz

Cichocki

et al. 2015

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

In this paper, we propose a method for solving the permutation problem that is inherent in the separation of convolved mixtures of speech signals in the time-frequency domain. The proposed method obtains the solution through maximization of a contrast function that exploits the similarity of the temporal envelope of the speech spectrum. For this purpose, the contrast calculation uses a global measure of similarity based on the recently developed family of generalized Alpha-Beta divergences, which depend on two tuning parameters, alpha and beta. This parameterization is exploited to best measure the similarity of the speech spectrum and to obtain solutions that are robust against noise and outliers. The ability of this contrast function to solve the permutation problem is supported by a theoretical study that shows that for a simple time-frequency speech model, the contrast value reaches its maximum when the estimated components are properly aligned. Several performance studies demonstrate that the proposed method maintains a high level of permutation correction accuracy in a wide variety of acoustic environments. Moreover, it produces better results than other state-of-the-art methods for solving permutations in highly reverberant environments.

show abstract

Section: Measuring the Similarity Of The Speech Spectrummentioning

confidence: 99%

“…The key assumption is that the probability density function (pdf) of each frequency bin is similar for all frequency components of a given speech source. Thus, the permutation can be corrected by considering small deviations in the parameters of the pdf model between neighboring bins [16] or by using several distance measures based on Information Theory [17].…”

mentioning

confidence: 99%

A Contrast Function Based on Generalized Divergences for Solving the Permutation Problem in Convolved Speech Mixtures

Sarmiento

Durán-Díaz

Cichocki

et al. 2015

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…From the constraint q + q 2 = 1, we obtain 2δ i q i = − q 2 , which implies that the higher-order terms in (10) are o( q 2 ) and can be neglected. Through a substitution, (10) can be simplified as…”

Section: Problem Formulationmentioning

confidence: 99%

“…Conventional BSS approaches do not ensure separation of source signals with a specified order in accordance with their stochastic properties. There are actually three types of source signals in [10,16]; they are known as standard normal, sup-Gaussian, and sub-Gaussian. The kurtosis in [27,32] is used to measure signals of the different distributions.…”

Section: Introductionmentioning

confidence: 99%

A Class of Sequential Blind Source Separation Method in Order Using Swarm Optimization Algorithm

Wang

Zhan

2015

Circuits Syst Signal Process

View full text Add to dashboard Cite

We consider the problem of sequential, blind source separation in some specific order from a mixture of sub-and sup-Gaussian sources. Three methods of separation are developed, specifically, kurtosis maximization using (a) particle swarm optimization, (b) differential evolution, and (c) artificial bee colony algorithm, all of which produce the separation in decreasing order of the absolute kurtosis based on the maximization of the kurtosis cost function. The validity of the methods was confirmed through simulation. Moreover, compared with other conventional methods, the proposed method separated the various sources with greater accuracy. Finally, we performed a real-world experiment to separate electroencephalogram (EEG) signals from a super-determined mixture with Gaussian noise. Whereas the conventional methods separate simultaneously EEG signals of interest along with noise, the result of this example shows the proposed methods recover from the outset solely those EEG signals of interest. This feature will be of benefit in many practical applications.

show abstract

“…The amplitude ambiguity is usually solved by the Minimal Distortion Principle [4]. For the permutation ambiguity, there have been several approaches with different degrees of success, using for example the correlation among envelopes [5] or power ratios [2], the generalized coherence function [6], the use of information theoretic distance measures [7], or the pseudoanechoic model for blind source separation [8]. In the last case, the problem was simplified in such a way that it required the robust estimation of the separation matrix for only one frequency bin, which then will serve to estimate the parameters for defining the separation matrices for all bins.…”

Section: Introductionmentioning

confidence: 99%

Using multiple frequency bins for stabilization of FD-ICA algorithms

Persia

Milone

2016

Signal Processing

View full text Add to dashboard Cite

a b s t r a c tIn the frequency domain independent component analysis approaches for audio sources separation, the convolutive mixing problem is replaced by the solution of several instantaneous mixing problems, one for each frequency bin of the short time Fourier transform. This methodology yields good results but requires the solution of the permutation ambiguity. Moreover, the performance of the separation algorithms for each bin is not guaranteed to be equivalent, thus some bins can have worse results than others. In this paper a technique based on data from multiple bins is proposed to address these issues. The use of multiple bin information produces a coupling of the separation, resulting in more stable separation matrices and reducing the occurrence of permutations, but increasing the computational cost. This can be mitigated by a sub-sampling of the multiple bins information. The results show that both approaches are beneficial for the frequency domain ICA approach, producing better separation in terms of objective quality measures.

show abstract

Using information theoretic distance measures for solving the permutation problem of blind source separation of speech signals

Cited by 11 publications

References 36 publications

A Contrast Function Based on Generalized Divergences for Solving the Permutation Problem in Convolved Speech Mixtures

A Contrast Function Based on Generalized Divergences for Solving the Permutation Problem in Convolved Speech Mixtures

A Class of Sequential Blind Source Separation Method in Order Using Swarm Optimization Algorithm

Using multiple frequency bins for stabilization of FD-ICA algorithms

Contact Info

Product

Resources

About