Underdetermined Blind Source Separation of Convolutive Mixtures by Hierarchical Clustering and L1-Norm Minimization

Winter, Stefan; Kellermann, Walter; Sawada, Hiroshi; Makino, Shoji

doi:10.1007/978-1-4020-6479-1_10

Cited by 25 publications

(52 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Under-determined ICA (N > M) is still difficult to solve, and we do not usually follow the above procedure, but directly estimate basis vectors a i (f ), as shown in e.g. [25].…”

Section: A Independent Component Analysis (Ica)mentioning

confidence: 99%

“…The authors are with NTT Communication Science Laboratories, NTT Corporation, 2-4 Hikaridai, Seika-cho, Soraku-gun, Kyoto 619-0237, Japan (e-mail: sawada@cslab.kecl.ntt.co.jp; shoko@cslab.kecl.ntt.co.jp; ryo@cslab.kecl.ntt.co.jp; maki@cslab.kecl.ntt.co.jp, phone: +81-774-93-5272, fax: +81-774-93-5158). EDICS: AUD-SSEN, AUD-LMAP An efficient and practical approach for such convolutive mixtures is frequency-domain BSS [7]- [25], where we apply a short-time Fourier transform (STFT) to the sensor observations x j (t). In the frequency domain, the convolutive mixture (1) can be approximated as an instantaneous mixture at each frequency:…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Grouping Separated Frequency Components by Estimating Propagation Model Parameters in Frequency-Domain Blind Source Separation

Sawada

Araki

Mukai

et al. 2007

IEEE Trans. Audio Speech Lang. Process.

Self Cite

109

View full text Add to dashboard Cite

Abstract-This paper proposes a new formulation and optimization procedure for grouping frequency components in frequency-domain blind source separation (BSS). We adopt two separation techniques, independent component analysis (ICA) and time-frequency (T-F) masking, for the frequency-domain BSS. With ICA, grouping the frequency components corresponds to aligning the permutation ambiguity of the ICA solution in each frequency bin. With T-F masking, grouping the frequency components corresponds to classifying sensor observations in the time-frequency domain for individual sources. The grouping procedure is based on estimating anechoic propagation model parameters by analyzing ICA results or sensor observations. More specifically, the time delays of arrival and attenuations from a source to all sensors are estimated for each source. The focus of this paper includes the applicability of the proposed procedure for a situation with wide sensor spacing where spatial aliasing may occur. Experimental results show that the proposed procedure effectively separates two or three sources with several sensor configurations in a real room, as long as the room reverberation is moderately low.

show abstract

“…Under-determined ICA (N > M) is still difficult to solve, and we do not usually follow the above procedure, but directly estimate basis vectors a i (f ), as shown in e.g. [25].…”

Section: A Independent Component Analysis (Ica)mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Grouping Separated Frequency Components by Estimating Propagation Model Parameters in Frequency-Domain Blind Source Separation

Sawada

Araki

Mukai

et al. 2007

IEEE Trans. Audio Speech Lang. Process.

Self Cite

109

View full text Add to dashboard Cite

show abstract

“…A range of values of p were explored in (Vincent, 2007). The value p=1 yielding the Laplacian distribution is often assumed (Winter, Kellermann, Sawada, & Makino, 2007). Alternative circular and non-circular sparse distributions, some of which were implicitly specified by their score function Φ(S jnf )=∂ log P(S jnf )/∂ S jnf , were used in (Smaragdis, 1998;Zibulevsky, Pearlmutter, Bofill, & Kisilev, 2001;Cemgil, Févotte, & Godsill, 2007).…”

Section: Continuous Local Time-frequency Linear Modelsmentioning

confidence: 99%

“…In the case when the number of sources is larger than the number of mixture channels, sparse component analysis (SCA) algorithms may be used instead. Most SCA algorithms also adopt approximate inference strategies, such as deriving the MAP source coefficients given fixed mixing vectors estimated via a binary model (Zibulevsky et al, 2001;Winter et al, 2007). When p is small, MAP inference under a generalized Gaussian prior splits the mixture in each time-frequency bin between the I sources whose DOAs are closest to the incoming sound direction and sets other source components to zero (Vincent, 2007).…”

Section: Continuous Local Time-frequency Linear Modelsmentioning

confidence: 99%

“…FDICA-based systems have been applied to the separation of live recordings of nonmoving sources (Smaragdis, 1998;Sawada et al, 2007;Nesta, Omologo, & Svaizer, 2008) and moving sources (Mukai, Sawada, Araki, & Makino, 2004), and SCA-based systems to the separation of instantaneous or echoic mixtures (Zibulevsky et al, 2001;Peterson & Kadambe, 2003;Cemgil et al, 2007;Vincent, 2007;Nesbit et al, 2009) and live recordings of nonmoving sources (Winter et al, 2007). These systems generally outperform binary or soft masking-based systems.…”

Section: Continuous Local Time-frequency Linear Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Probabilistic Modeling Paradigms for Audio Source Separation

et al.

View full text Add to dashboard Cite

Most sound scenes result from the superposition of several sources, which can be separately perceived and analyzed by human listeners. Source separation aims to provide machine listeners with similar skills by extracting the sounds of individual sources from a given scene. Existing separation systems operate either by emulating the human auditory system or by inferring the parameters of probabilistic sound models. In this chapter, we focus on the latter approach and provide a joint overview of established and recent models, including independent component analysis, local time-frequency models and spectral template-based models. We show that most models are instances of one of the following two general paradigms: linear modeling or variance modeling. We compare the merits of either paradigm and report objective performance figures. We conclude by discussing promising combinations of probabilistic priors and inference algorithms that could form the basis of future state-of-the-art systems.

show abstract

K-means Based Underdetermined Blind Speech Separation

Araki

Sawada

Makino

2007

Blind Speech Separation

Self Cite

View full text Add to dashboard Cite

Abstract. This chapter addresses a blind sparse source separation method that can employ arbitrarily arranged multiple microphones. Some sparse source separation methods, which rely on source sparseness and an anechoic mixing model, have already been proposed. The validity of the sparseness and anechoic assumptions will be investigated in this chapter. As most of the existing methods utilize a stereo (two sensors) system, they limit the separation ability to a 2-dimensional half-plane. This chapter describes a method for multiple microphones. This method employs the k-means algorithm, which is an efficient clustering algorithm. The method can be easily applied to three or more sensors arranged nonlinearly. Promising results were obtained for 2-and 3-dimensionally distributed speech signals with nonlinear/nonuniform sensor arrays in a real room even in underdetermined situations.

show abstract

Underdetermined Blind Source Separation of Convolutive Mixtures by Hierarchical Clustering and L1-Norm Minimization

Cited by 25 publications

References 53 publications

Grouping Separated Frequency Components by Estimating Propagation Model Parameters in Frequency-Domain Blind Source Separation

Grouping Separated Frequency Components by Estimating Propagation Model Parameters in Frequency-Domain Blind Source Separation

Probabilistic Modeling Paradigms for Audio Source Separation

K-means Based Underdetermined Blind Speech Separation

Contact Info

Product

Resources

About