Search

Showing total 2 results
Start Over Topic noise Publication Year Range Last 50 years
2 results

1. Explaining Deep Learning Models for Speech Enhancement

Author

Sunit Sivasankaran, Emmanuel Vincent, Dominique Fohr, Microsoft Corporation [Redmond], Microsoft Corporation [Redmond, Wash.], Speech Modeling for Facilitating Oral-Based Communication (MULTISPEECH), Inria Nancy - Grand Est, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Department of Natural Language Processing & Knowledge Discovery (LORIA - NLPKD), Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL), This work was made with the support of the French National Research Agency, in the framework of the project VOCADOM 'Robust voice command adapted to the user and to the context for AAL' (ANR-16-CE33-0006). Experiments presented in this paper were carried out using the Grid’5000 testbed, supported by a scientific interest group hosted by Inria and including CNRS, RENATER and several Universities as well as otherorganizations (see https://www.grid5000.fr)., ANR-16-CE33-0006,VOCADOM,Commande vocale robuste adaptée à la personne et au contexte pour l'autonomie à domicile(2016), Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), and Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)

Subjects
Artificial neural network, business.industry, Computer science, Deep learning, Speech recognition, Word error rate, 020206 networking & telecommunications, Context (language use), 02 engineering and technology, explainable AI, Speech enhancement, 030507 speech-language pathology & audiology, 03 medical and health sciences, Noise, feature attribution, [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG], Robustness (computer science), [INFO.INFO-SD]Computer Science [cs]/Sound [cs.SD], 0202 electrical engineering, electronic engineering, information engineering, Feature (machine learning), speech enhancement, Artificial intelligence, 0305 other medical science, business
Abstract

International audience; We consider the problem of explaining the robustness of neural networks used to compute time-frequency masks for speech enhancement to mismatched noise conditions. We employ the Deep SHapley Additive exPlanations (DeepSHAP) feature attribution method to quantify the contribution of every timefrequency bin in the input noisy speech signal to every timefrequency bin in the output time-frequency mask. We define an objective metric-referred to as the speech relevance scorethat summarizes the obtained SHAP values and show that it correlates with the enhancement performance, as measured by the word error rate on the CHiME-4 real evaluation dataset. We use the speech relevance score to explain the generalization ability of three speech enhancement models trained using synthetically generated speech-shaped noise, noise from a professional sound effects library, or real CHiME-4 noise. To the best of our knowledge, this is the first study on neural network explainability in the context of speech enhancement.

Published
2021

2. A benchmark of kriging-based infill criteria for noisy optimization

Author

Tobias Wagner, David Ginsbourger, Victor Picheny, Unité de Mathématiques et Informatique Appliquées de Toulouse (MIAT INRA), Institut National de la Recherche Agronomique (INRA), Technische Universität Dortmund [Dortmund] (TU), University of Bern, Deutsche Forschungsgemeinschaft (DFG) [SFB/TR TRR 30] Funding Text : The contributions of Tobias Wagner to this paper are based on investigations of the project D5 of the Collaborative Research Center SFB/TR TRR 30, which is kindly supported by the Deutsche Forschungsgemeinschaft (DFG)., and Picheny, Victor

Subjects
Mathematical optimization, Control and Optimization, Covariance function, 0211 other engineering and technologies, 02 engineering and technology, Metamodeling, symbols.namesake, 510 Mathematics, DESIGN, Kriging, Homoscedasticity, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], [MATH]Mathematics [math], Gaussian process, GLOBAL OPTIMIZATION, Mathematics, EGO, 021103 operations research, [MATH.MATH-OC] Mathematics [math]/Optimization and Control [math.OC], Covariance, Computer Graphics and Computer-Aided Design, Computer Science Applications, COMPUTER EXPERIMENTS, Control and Systems Engineering, Gaussian noise, SIMULATION, symbols, 020201 artificial intelligence & image processing, Engineering design process, Noise, Software
Abstract

International audience; Responses of many real-world problems can only be evaluated perturbed by noise. In order to make an efficient optimization of these problems possible, intelligent optimization strategies successfully coping with noisy evaluations are required. In this article, a comprehensive review of existing kriging-based methods for the optimization of noisy functions is provided. In summary, ten methods for choosing the sequential samples are described using a unified formalism. They are compared on analytical benchmark problems, whereby the usual assumption of homoscedastic Gaussian noise made in the underlying models is meet. Different problem configurations (noise level, maximum number of observations, initial number of observations) and setups (covariance functions, budget, initial sample size) are considered. It is found that the choices of the initial sample size and the covariance function are not critical. The choice of the method, however, can result in significant differences in the performance. In particular, the three most intuitive criteria are found as poor alternatives. Although no criterion is found consistently more efficient than the others, two specialized methods appear more robust on average.

Published
2013