Skip to content
Free planNo credit card requiredStart writing
AI for Research home
AI for Research home
Dashboard
Sign inGet started

Search 250 million research papers

Read the abstract, see who wrote it and who cites it, and get the citation in APA, MLA, Chicago or Harvard.

About 10,000 papers for “Source separation”

Source Separation

Nikolaos Mitianoudis, Senior Member · 2011

Directional or Circular statistics are pertaining to the analysis and interpretation of directions or rotations. In this work, a novel probability distribution is proposed to model multidimensional sparse directional data. The Generalised Directional Laplacian Distribution (DLD) is a hybrid between the Lapla-cian distribution and the von Mises-Fisher distribution. The distribution’s parameters are estimated using Maximum-Likelihood Estimation over a set of training data points. Mixtures of Directional Laplacian Distributions (MDLD) are also introduced in order to model multiple concentrations of sparse directional data. The author explores the application of the derived DLD mixture model to cluster sound sources that exist in an underdetermined instantaneous sound mixture. The proposed model can solve the general K × L (K < L) underdetermined instantaneous source separation problem, offering a fast and stable solution.

Cited by 0

Source Separation

Jonathon Chambers, Wenwu Wang, An Example · 2016

• Introduce the source separation problem and its application domains • Key books and literature reviews • Technical preliminaries • Concepts of ICA – independence and non-Gaussianity

Cited by 0

Source separation

Vincent, Emmanuel · HAL (Le Centre pour la Communication Scientifique Directe) · 2013

Journées Thématiques 'Localisation, séparation et suivi de sources audio'

Cited by 0

Source control, source separation

Adriano Joss, Ursula Klaschka, Thomas Knacker et al. · DORA Eawag (Swiss Federal Institute of Aquatic Science and Technology (Eawag)) · 2006

The reduction of the micropollutant emission into the environment is to be achieved by complementing end-of-pipe advanced wastewater treatment steps in centralized municipal plants with measures taken at the source. A significant reduction in terms of consumed quantities as well as in terms of the ecotoxicity of the selected compounds is feasible if the stakeholders can be successfully involved:<br />\n• patients and the general public represent the starting point on which a shift of production (i.e. towards more environmentally friendly compounds) as well as legislative issues are based on<br />\n• industry is a central player able to influence the pace at which new solutions are introduced on the market<br />\n• physicians and health care personnel play a key role in the therapeutic choice as well as in supporting complementary strategies to reduce medical consumption (e.g. hygienic standard in hospitals, empirical vs. targeted therapy)<br />\n• legislative authorities are required if solutions are to be implemented beyond voluntary choice (e.g. composition of industrial or hospital wastewater discharged into the sewer)<br />\nEcolabelling may represent a tool to initiate the required public discourse for phasing in of new products or production methods. However, it ought to be substituted or backed up by legislation to achieve broader compliance or general validity. (...)

Cited by 3

Audio source separation

Makino, Shoji · CERN Document Server (European Organization for Nuclear Research) · 2018

Cited by 0

Sound Source Separation

Gianpaolo Evangelista, Sylvain Marchand, Mark D. Plumbley et al. · 2011

This chapter contains sections titled: Introduction Binaural source separation Source separation from single-channel signals Applications Conclusions Acknowledgements References

Cited by 4

Nonstationary source separation

Seungjin Choi, Oyoung Lee · 2003

Source separation is a statistical method, the goal of which is to recover mutually independent sources from their linear instantaneous mixtures without resorting to any prior knowledge. Most existing methods have been focused on stationary sources, so higher-order statistics was necessary for separation, unless sources are temporally correlated. For nonstationary sources, however, it was shown by Matsuoka, Ohya and Kawamoto (1995) that source separation could be achieved by only decorrelation (second-order statistics). In present paper, we adopt the natural gradient method of Amari (see Neural Computation, vol.10, p.251-76, 1998) and derive an efficient source separation algorithm by minimizing the cost function proposed by Matsuoka et al. The useful behavior of the proposed algorithm is demonstrated through computer simulations.

Cited by 16

Blind source separation

Miloud Frikel, V. Barroso, João Xavier · The Journal of the Acoustical Society of America · 1999

In signal processing the received data can be modeled as a superposition of a finite number of elementary source signals and an additive noise. Generally, in a multi-sensor environment application, such as underwater acoustics, the principal objective is estimating the number and directions of radiating sources. In the last decade, eigenstructure-based methods yielding resolution have been proposed. However, these methods need the model of sources. In many cases, this model is not available, indeed, there is the appearance of the blind methods. Blind technique refers to lack of knowledge about the mixing system. It assumes very weak assumptions either for signals or mixtures. Blind source separation consists in identifying the sensors/sources transfer matrix and retrieving the source signals without resorting a priori information about the above matrix. It exploits only the information carried by the received signals themselves. The potential benefit of such a blind approach is that source separation is essentially unaffected by errors in the propagation model or in array calibration whereas the classical array processing methods assume that the array geometry is known and the sources are in the far field of the array. This contribution deals with blind source separation using second-order statistics. [This work was partially funded by FEDER and PRAXIS XXI, under Contract No. PRAXIS/3/3.1/TPR/23/94.]

Cited by 3

UNISON SOURCE SEPARATION

Fabian-Robert Stöter, Stefan Bayer, Bernd Edler · 2014

In this work we present a new scenario of analyzing and separating linear mixtures of musical instrument signals. When instruments are playing in unison, traditional source separation methods are not performing well. Although the sources share the same pitch, they often still differ in their modulation frequency caused by vibrato and/or tremolo effects. In this paper we propose source separation schemes that exploit AM/FM characteristics to improve the sep-aration quality of such mixtures. We show a method to process mixtures based on differences in their amplitude modulation fre-quency of the sources by using non-negative tensor factorization. Further, we propose an informed warped time domain approach for separating mixtures based on variations in the instantaneous frequencies of the sources. 1.

Cited by 7

Bayesian Source Separation

Saïd Moussaoui, Leonardo Tomazeli Duarte, Nicolas Dobigeon et al. · 2023

Solving a Source separation problem using a maximum likelihood approach offers the possibility to encode, in addition to the mutual statistical independence of the sources, additional prior information on these signals by specifying their probability distribution functions. Such setting also corresponds to some specific choice of the non-linear functions in independent component analysis (ICA) algorithms based on non-linear decorrelation (See chapter \ref{chap:1} of this book). The Bayesian inference strategy offers an additional flexibility by allowing to take into account the noise statistics and to account for prior information on the mixing coefficients. The purpose of this chapter is to present the Bayesian approach for source separation. The general framework of Bayesian estimation will be presented in the first section of this chapter. It includes the specification of the likelihood resulting from the statistical description of the noise and the formulation of statistical models encoding the available information on the sought source signals and mixing coefficients. Resulting algorithms in the case of linear or nonlinear mixing models in the context of physical-chemical sensing will be presented. The Bayesian approach will be illustrated through some examples of case studies based on spectral data resulting from spectrometry measurements.

Cited by 0Open access

Musical Source Separation

Mocanu, Alexandru · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 2020

Musical source separation is a complex topic that has been extensively explored in the signal processing community and has benefited greatly from recent machine learning research. Many deep learning models with impressive source separation quality have been released in the last couple of years, all of them dealing with studio recorded music split into four instrument categories, vocals, drums, bass and other. We study how we can extend the number of instrument categories and conclude that electric guitar is also feasible to separate. We then turn our attention towards learning relevant signal encodings using parameterized filterbanks and we observe that filterbanks can not improve over simple convolutions on their own, but can help if the encoder is composed of both convolutions and filterbanks. Finally, we try to adapt models trained on studio music to live music separation and conclude that models trained on clean data also provide the best performance on live music as well.

Cited by 0Open access

Music Source Separation

Holík, Viliam · Brno University of Technology Digital Library (Brno University of Technology) · 2023

Na separáciu zdrojov z hudobných nahrávok sa používajú neurónové siete. Jednou z takýchto sietí je Conv-TasNet. Cieľom práce je experimentovať s už existujúcou implementáciou tejto siete za účelom potenciálneho zlepšenia. Trénovanie modelov prebiehalo na dátovej sade MUSDB18. Postupne sa experimentovalo so zmenou štruktúry siete, transformáciou signálov z časovej domény do frekvenčnej pre účely počítania objektívnej funkcie, zámenou rôznych objektívnych funkcií za pôvodnú, hľadaním optimálneho koeficientu rýchlosti učenia pre každú objektívnu funkciu a jeho postupným zmenšovaním v priebehu učenia. Ako najlepšie experimenty podla metriky SDR vyšli trénovania s objektívnymi funkciami L1 a logaritmickou L2 v časovej doméne pri vyššom počiatočnom koeficiente rýchlosti učenia s jeho postupným zmenšovaním v priebehu učenia. V relatívnom porovnaní najlepších modelov oproti východziemu ide o viac ako 2,5% zlepšenie.

Cited by 0