Generalized independent low-rank matrix analysis using heavy-tailed distributions for blind source separation

In this paper, we propose new accelerated update rules for rank-constrained spatial covariance model estimation, which efficiently extracts a directional target source in diffuse background noise. The naive update rule requires heavy computation such as matrix inversion or matrix multiplication. We resolve this problem by expanding matrix inversion to reduce computational complexity; in the parameter update step, we need neither matrix inversion nor multiplication. In an experiment, we show that the proposed accelerated update rule achieves 87 times faster calculation than the naive one.

show abstract

“…ij for all i and j; 8λ i ← λ i for all i; 9 Calculateγ ij by (34) and (50) for all i and j; 10 Update r…”

Section: Proposed Methods a Motivationmentioning

confidence: 99%

Acceleration of rank-constrained spatial covariance matrix estimation for blind speech extraction

Kubo

Takamune

Kitamura

et al. 2019

2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)

Self Cite

View full text Add to dashboard Cite

show abstract

“…In GGD-ILRMA, the cost function (9) is minimized by alternately repeating the update of the demixing matrix W i using (13)- (15) and the update of the low-rank source models T n and V n using (16) and (17), respectively. A monotonic decrease in the cost is guaranteed over these update rules.…”

Section: B Update Rule For Low-rank Source Modelmentioning

confidence: 99%

“…In GGD-ILRMA with β = 4, the demixing matrix W i is updated by (58)-(64), and the low-rank models T n and V n are updated by (16) and (17), respectively. These update rules are derived using the MM algorithm and GIP-HSM, thus guaranteeing a monotonic decrease of the cost function (30).…”

Section: Sub-gaussian Ilrma Based On Gip-hsmmentioning

confidence: 99%

“…We compared the separation performance of the proposed sub-Gaussian GGD-ILRMA (β = 4) with those of conventional IS-ILRMA [9] and GGD-ILRMA (β < 2) [16]. We artificially produced monaural dry music sources of four melody parts (melody 1: main melody, melody 2: counter melody, midrange, and bass) using Microsoft GS Wavetable Synth, where several musical instruments were chosen to play these melody parts [20], [21].…”

Section: A Bss Experiments On Music Signalsmentioning

confidence: 99%

“…Recently, various types of source generative model have been proposed in ILRMA for robust BSS. In particular, t-ILRMA [15] and GGD-ILRMA [16], [17] have been proposed as generalizations of IS-ILRMA with a complex Student's t distribution and a complex generalized Gaussian distribution (GGD), respectively. In t-ILRMA and GGD-ILRMA, the kurtosis of the generative models' distributions can be parametrically changed along with the degree-of-freedom parameter in Student's t distribution and the shape parameter in the GGD.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Independent Low-Rank Matrix Analysis Based on Time-Variant Sub-Gaussian Source Model

Mogami

Takamune

Kitamura

et al. 2018

2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)

Self Cite

View full text Add to dashboard Cite

Independent low-rank matrix analysis (ILRMA) is a fast and stable method for blind audio source separation. Conventional ILRMAs assume time-variant (super-)Gaussian source models, which can only represent signals that follow a super-Gaussian distribution. In this paper, we focus on ILRMA based on a generalized Gaussian distribution (GGD-ILRMA) and propose a new type of GGD-ILRMA that adopts a timevariant sub-Gaussian distribution for the source model. By using a new update scheme called generalized iterative projection for homogeneous source models, we obtain a convergence-guaranteed update rule for demixing spatial parameters. In the experimental evaluation, we show the versatility of the proposed method, i.e., the proposed time-variant sub-Gaussian source model can be applied to various types of source signal.

show abstract

Independent deeply learned matrix analysis with automatic selection of stable microphone-wise update and fast sourcewise update of demixing matrix

et al. 2021

View full text Add to dashboard Cite

Generalized independent low-rank matrix analysis using heavy-tailed distributions for blind source separation

Cited by 46 publications

References 31 publications

Acceleration of rank-constrained spatial covariance matrix estimation for blind speech extraction

Acceleration of rank-constrained spatial covariance matrix estimation for blind speech extraction

Independent Low-Rank Matrix Analysis Based on Time-Variant Sub-Gaussian Source Model

Independent deeply learned matrix analysis with automatic selection of stable microphone-wise update and fast sourcewise update of demixing matrix

Contact Info

Product

Resources

About