1 resultado para initialization uncertainty
em Bucknell University Digital Commons - Pensilvania - USA
Filtro por publicador
- Repository Napier (1)
- Aberdeen University (5)
- Aberystwyth University Repository - Reino Unido (1)
- Acceda, el repositorio institucional de la Universidad de Las Palmas de Gran Canaria. España (4)
- AMS Tesi di Dottorato - Alm@DL - Università di Bologna (4)
- Aquatic Commons (5)
- Archimer: Archive de l'Institut francais de recherche pour l'exploitation de la mer (1)
- Archive of European Integration (4)
- Archivo Digital para la Docencia y la Investigación - Repositorio Institucional de la Universidad del País Vasco (8)
- Aston University Research Archive (61)
- Biblioteca Digital da Produção Intelectual da Universidade de São Paulo (4)
- Biblioteca Digital da Produção Intelectual da Universidade de São Paulo (BDPI/USP) (2)
- BORIS: Bern Open Repository and Information System - Berna - Suiça (32)
- Boston University Digital Common (1)
- Brock University, Canada (2)
- Bucknell University Digital Commons - Pensilvania - USA (1)
- Bulgarian Digital Mathematics Library at IMI-BAS (3)
- CaltechTHESIS (4)
- Cambridge University Engineering Department Publications Database (114)
- CentAUR: Central Archive University of Reading - UK (122)
- Chinese Academy of Sciences Institutional Repositories Grid Portal (12)
- Cochin University of Science & Technology (CUSAT), India (1)
- Collection Of Biostatistics Research Archive (4)
- Comissão Econômica para a América Latina e o Caribe (CEPAL) (3)
- CORA - Cork Open Research Archive - University College Cork - Ireland (4)
- Corvinus Research Archive - The institutional repository for the Corvinus University of Budapest (6)
- CUNY Academic Works (4)
- Dalarna University College Electronic Archive (2)
- Deakin Research Online - Australia (52)
- Department of Computer Science E-Repository - King's College London, Strand, London (1)
- Digital Commons @ DU | University of Denver Research (1)
- Digital Commons at Florida International University (7)
- Digital Peer Publishing (1)
- DigitalCommons@The Texas Medical Center (4)
- DigitalCommons@University of Nebraska - Lincoln (2)
- Doria (National Library of Finland DSpace Services) - National Library of Finland, Finland (2)
- DRUM (Digital Repository at the University of Maryland) (3)
- Duke University (2)
- Earth Simulator Research Results Repository (1)
- eResearch Archive - Queensland Department of Agriculture; Fisheries and Forestry (2)
- Glasgow Theses Service (2)
- Greenwich Academic Literature Archive - UK (3)
- Harvard University (1)
- Helda - Digital Repository of University of Helsinki (10)
- Indian Institute of Science - Bangalore - Índia (41)
- INSTITUTO DE PESQUISAS ENERGÉTICAS E NUCLEARES (IPEN) - Repositório Digital da Produção Técnico Científica - BibliotecaTerezine Arantes Ferra (3)
- Instituto Politécnico do Porto, Portugal (10)
- Massachusetts Institute of Technology (4)
- Memorial University Research Repository (1)
- Ministerio de Cultura, Spain (3)
- Nottingham eTheses (1)
- Plymouth Marine Science Electronic Archive (PlyMSEA) (7)
- Publishing Network for Geoscientific & Environmental Data (4)
- QSpace: Queen's University - Canada (3)
- QUB Research Portal - Research Directory and Institutional Repository for Queen's University Belfast (62)
- Queensland University of Technology - ePrints Archive (85)
- Repositório Científico da Universidade de Évora - Portugal (1)
- Repositório Científico do Instituto Politécnico de Lisboa - Portugal (1)
- Repositório digital da Fundação Getúlio Vargas - FGV (21)
- Repositorio Institucional de la Universidad de Málaga (2)
- Repositório Institucional UNESP - Universidade Estadual Paulista "Julio de Mesquita Filho" (16)
- Research Open Access Repository of the University of East London. (2)
- RUN (Repositório da Universidade Nova de Lisboa) - FCT (Faculdade de Cienecias e Technologia), Universidade Nova de Lisboa (UNL), Portugal (1)
- SAPIENTIA - Universidade do Algarve - Portugal (1)
- School of Medicine, Washington University, United States (1)
- Universidad de Alicante (6)
- Universidad del Rosario, Colombia (2)
- Universidad Politécnica de Madrid (44)
- Universitat de Girona, Spain (4)
- Universitätsbibliothek Kassel, Universität Kassel, Germany (3)
- Université de Montréal, Canada (8)
- Université Laval Mémoires et thèses électroniques (1)
- University of Canberra Research Repository - Australia (1)
- University of Connecticut - USA (1)
- University of Michigan (10)
- University of Queensland eSpace - Australia (41)
- University of Southampton, United Kingdom (1)
- University of Washington (5)
- WestminsterResearch - UK (2)
Resumo:
We present a new approach for corpus-based speech enhancement that significantly improves over a method published by Xiao and Nickel in 2010. Corpus-based enhancement systems do not merely filter an incoming noisy signal, but resynthesize its speech content via an inventory of pre-recorded clean signals. The goal of the procedure is to perceptually improve the sound of speech signals in background noise. The proposed new method modifies Xiao's method in four significant ways. Firstly, it employs a Gaussian mixture model (GMM) instead of a vector quantizer in the phoneme recognition front-end. Secondly, the state decoding of the recognition stage is supported with an uncertainty modeling technique. With the GMM and the uncertainty modeling it is possible to eliminate the need for noise dependent system training. Thirdly, the post-processing of the original method via sinusoidal modeling is replaced with a powerful cepstral smoothing operation. And lastly, due to the improvements of these modifications, it is possible to extend the operational bandwidth of the procedure from 4 kHz to 8 kHz. The performance of the proposed method was evaluated across different noise types and different signal-to-noise ratios. The new method was able to significantly outperform traditional methods, including the one by Xiao and Nickel, in terms of PESQ scores and other objective quality measures. Results of subjective CMOS tests over a smaller set of test samples support our claims.