Auxiliary-Function-Based Independent Component Analysis for Super-Gaussian Sources

Ono, Nobutaka; Miyabe, Shigeki

doi:10.1007/978-3-642-15995-4_21

Cited by 86 publications

(88 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These methods exploit the statistical independence between specific sources and estimate a demixing matrix for the separation. For both ICA and IVA, fast and stable update rules, which are derived by an auxiliary function technique, have been proposed [11], [12].…”

Section: Introductionmentioning

confidence: 99%

Experimental analysis of optimal window length for independent low-rank matrix analysis

Kitamura

Saruwatari

2017

2017 25th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

Abstract-In this paper, we address the blind source separation (BSS) problem and analyze the optimal window length in the short-time Fourier transform (STFT) for independent low-rank matrix analysis (ILRMA). ILRMA is a state-of-the-art BSS technique that utilizes the statistical independence between low-rank matrix spectrogram models, which are estimated by nonnegative matrix factorization. In conventional frequency-domain BSS, the modeling error of a mixing system increases when the window length is too short, and the accuracy of statistical estimation decreases when the window length is too long. Therefore, the optimal window length is determined by both the reverberation time and the number of time frames. However, unlike classical BSS methods such as ICA and IVA, ILRMA enables the full modeling of spectrograms, which may improve the robustness to a decrease in the number of frames in a longer-window case. To confirm this hypothesis, the optimal window length for ILRMA is experimentally investigated, and the difference between the performances of ILRMA and conventional BSS is discussed.

show abstract

Section: Introductionmentioning

confidence: 99%

Experimental analysis of optimal window length for independent low-rank matrix analysis

Kitamura

Saruwatari

2017

2017 25th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

show abstract

“…2) M-step of EM algorithm: In the M-step, we update the model parameters according to the problem (18). The update rules for {π n,t,c } n,t,c are easily obtained as…”

Section: ) E-step Of Em Algorithmmentioning

confidence: 99%

“…It will sometimes be abbreviated as g c (r n,F,t ). Suppose also that g c satisfies the following two conditions as in [16]- [18]: Under the above assumptions, we can derive the following auxiliary function J(Θ) of the cost (18) in the same way as described in [16]- [18]: (22) where C is independent of {w n,f } n,f , and…”

Section: ) E-step Of Em Algorithmmentioning

confidence: 99%

“…On the other hand, in the case where F = {{f } : f ∈ [N F ]} it is nothing but the conventional (auxiliary function based) FD-ICA [4], [18]. In this sense, the proposed method is regarded as an extension of FD-ICA and IVA.…”

Section: Relation To Prior Workmentioning

confidence: 99%

“…Besides developing the new model, we derive an optimization algorithm for it, where the states of sources, IVA filters, and permutation alignments between each pair of frequency bands are jointly optimized by combining an EM algorithm with a fast and stable auxiliary function method for IVA proposed in [16]- [18]. Owing to the simultaneous optimization, the proposed algorithm can robustly fit the prior of the model to the distribution of each source.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Independent vector analysis with frequency range division and prior switching

Ikeshita

Kawaguchi

Togami

et al. 2017

2017 25th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

Abstract-A novel source model is developed to improve the separation performance of independent vector analysis (IVA) for speech mixtures. The source model of IVA generally assumes the same amount of statistical dependency on each pair of frequency bins, which is not effective for speech signals with strong correlations among neighboring frequency bins. In the proposed model, the set of all frequency bins is divided into frequency bands, and the statistical dependency is assumed only within each band to better represent speech signals. In addition, each source prior is switched depending on the source states, active or inactive, since intermittent silent periods have totally different priors from those of speech periods. The optimization of the model is based on an EM algorithm, in which the IVA filters, states of sources, and permutation alignments between each pair of bands are jointly optimized. The experimental results show the effectiveness of the proposed model.Index Terms-Blind source separation, independent vector analysis (IVA), independent component analysis (ICA)

show abstract

A nonparametric Bayesian model for system identification based on a super‐Gaussian distribution

Tanji

Murakami

Kamata

2019

Elect Comm in Japan

View full text Add to dashboard Cite

In the acoustic signal processing applications of finite impulse response (FIR) system identification, it is important to develop the identification method that is robust to super‐Gaussian noises. Moreover, the identification method that estimates the FIR coefficients and the order of the unknown system is required, because the order of the unknown system is unavailable in advance. Therefore, in this paper, we propose a nonparametric Bayesian (NPB) model for FIR system identification using a super‐Gaussian likelihood and the beta‐Bernoulli process. In the proposed NPB model, we employ the hyperbolic secant distribution for the likelihood function. Then, we derive the inference algorithm to simultaneously estimate the FIR coefficients and the order of the unknown system. Our inference algorithm based on a hybrid inference approach combining the majorization‐minimization (MM) algorithm and the Gibbs sampler. The simulation results suggest that the proposed method outperforms the conventional identification algorithms in a super‐Gaussian noise environment.

show abstract

Auxiliary-Function-Based Independent Component Analysis for Super-Gaussian Sources

Cited by 86 publications

References 7 publications

Experimental analysis of optimal window length for independent low-rank matrix analysis

Experimental analysis of optimal window length for independent low-rank matrix analysis

Independent vector analysis with frequency range division and prior switching

A nonparametric Bayesian model for system identification based on a super‐Gaussian distribution

Contact Info

Product

Resources

About