Variational EM for binaural sound-source separation and localization

Deleforge, Antoine; Forbes, Florence; Horaud, Radu

doi:10.1109/icassp.2013.6637612

Cited by 24 publications

(32 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the arXiv:1610.04770v1 [cs.SD] 15 Oct 2016 binaural hearing context, Deleforge and Horaud have proposed a probabilistic piecewise affine regression model that infers the localization-to-interaural data mapping and its inverse [18]. They have extended this approach to the case of multiple sources using the variational Expectation Maximization (EM) framework [19], [20]. In [21], another approach was presented based on a Gaussian Mixture Model (GMM) which was used to learn the azimuth-dependent distribution of the binaural feature space.…”

Section: Introductionmentioning

confidence: 99%

Semi-Supervised Source Localization on Multiple Manifolds With Distributed Microphones

Laufer-Goldshtein

Talmon

Gannot

2017

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

The problem of source localization with ad hoc microphone networks in noisy and reverberant enclosures, given a training set of prerecorded measurements, is addressed in this paper. The training set is assumed to consist of a limited number of labelled measurements, attached with corresponding positions, and a larger amount of unlabelled measurements from unknown locations. However, microphone calibration is not required. We use a Bayesian inference approach for estimating a function that maps measurement-based feature vectors to the corresponding positions. The central issue is how to combine the information provided by the different microphones in a unified statistical framework. To address this challenge, we model this function using a Gaussian process with a covariance function that encapsulates both the connections between pairs of microphones and the relations among the samples in the training set. The parameters of the process are estimated by optimizing a maximum likelihood (ML) criterion. In addition, a recursive adaptation mechanism is derived where the new streaming measurements are used to update the model. Performance is demonstrated for 2-D localization of both simulated data and real-life recordings in a variety of reverberation and noise levels.

show abstract

Section: Introductionmentioning

confidence: 99%

Semi-Supervised Source Localization on Multiple Manifolds With Distributed Microphones

Laufer-Goldshtein

Talmon

Gannot

2017

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…The second direction is data-driven, and consists in learning a mapping from measured high-dimensional acoustic features to source postions. Such mappings are learned from carefully recorded datasets in a supervised [5,6] or semi-supervised [7] way. Since obtaining these datasets is time consuming, the methods are usually working well for one specific room and setup, and are hard to generalize in practice.…”

Section: Introductionmentioning

confidence: 99%

Hearing in a shoe-box: Binaural source position and wall absorption estimation using virtually supervised learning

Kataria

Gaultier

Deleforge

2017

2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

This paper introduces a new framework for supervised sound source localization referred to as virtually-supervised learning. An acoustic shoe-box room simulator is used to generate a large number of binaural single-source audio scenes. These scenes are used to build a dataset of spatial binaural features annotated with acoustic properties such as the 3D source position and the walls' absorption coefficients. A probabilistic high-to low-dimensional regression framework is used to learn a mapping from these features to the acoustic properties. Results indicate that this mapping successfully estimates the azimuth and elevation of new sources, but also their range and even the walls' absorption coefficients solely based on binaural signals. Results also reveal that incorporating randomdiffusion effects in the data significantly improves the estimation of all parameters.

show abstract

“…Initializing˜ to zero and choosing permutation π, in the first step, the squared distance matrix M˜ π as defined in (10) is double centered [28] 3 (steps 4) followed by SVD to obtain a low-rank matrixM˜ π along with the position matrixX˜ π ;d π is equal to the last row ofX˜ π . In the second step,˜ is updated by solving (11) and (12). Based on the new estimate of˜ ,M˜ π is updated using (10).…”

Section: ) Synchronizationmentioning

confidence: 99%

“…Furthermore, the data-driven learning and generative modeling of location-dependent spatial characteristics has been shown promising for sound source localization in a reverberant environment; in [12] and [13] room-and microphone locationspecific models were trained on white noise signals and incorporated for 2D-localization with two microphones. Nesta and Omologo [14] presented an approach that exploited sparsity of source signals in the cross-power spectral domain and accounted in a statistical manner for deviations of the sources' spatial characteristics from an ideal anechoic propagation model caused by multipath effect.…”

Section: Introductionmentioning

confidence: 99%

Spatial Sound Localization via Multipath Euclidean Distance Matrix Recovery

Taghizadeh

Asaei

Haghighatshoar

et al. 2015

IEEE J. Sel. Top. Signal Process.

View full text Add to dashboard Cite

Abstract-A novel localization approach is proposed in order to find the position of an individual source using recordings of a single microphone in a reverberant enclosure. The multipath propagation is modeled by multiple virtual microphones as images of the actual single microphone and a multipath distance matrix is constructed whose components consist of the squared distances between the pairs of microphones (real or virtual) or the squared distances between the microphones and the source. The distances between the actual and virtual microphones are computed from the geometry of the enclosure. The microphonesource distances correspond to the support of the early reflections in the room impulse response associated with the source signal acquisition. The low-rank property of the Euclidean distance matrix is exploited to identify this correspondence. Source localization is achieved through optimizing the location of the source matching those measurements. The recording time of the microphone and generation of the source signal is asynchronous and estimated via the proposed procedure. Furthermore, a theoretically optimal joint localization and synchronization algorithm is derived by formulating the source localization as minimization of a quartic cost function. It is shown that the global minimum of the proposed cost function can be efficiently computed by converting it to a generalized trust region sub-problem. Numerical simulations on synthetic data and real data recordings obtained by practical tests show the effectiveness of the proposed approach.

show abstract

Variational EM for binaural sound-source separation and localization

Cited by 24 publications

References 17 publications

Semi-Supervised Source Localization on Multiple Manifolds With Distributed Microphones

Semi-Supervised Source Localization on Multiple Manifolds With Distributed Microphones

Hearing in a shoe-box: Binaural source position and wall absorption estimation using virtually supervised learning

Spatial Sound Localization via Multipath Euclidean Distance Matrix Recovery

Contact Info

Product

Resources

About