Missing feature reconstruction methods for robust speaker identification. Zhang, X., Zhang, H., & Gao, G. In 2014 22nd European Signal Processing Conference (EUSIPCO), pages 1482-1486, Sep., 2014. Paper abstract bibtex In this study, we propose a reconstruction method to restore the degraded features for robust speaker identification. The proposed method is based on a hybrid generative model which consists of deep belief network (DBN) and restricted Boltzmann machine (RBM). Specifically, the noisy speech is firstly decomposed into time-frequency (T-F) representations. Then ideal binary mask (IBM) is computed to indicate each T-F point as reliable or unreliable. We reconstruct the unreliable ones by the proposed model iteratively. Finally, reconstructed feature is utilized to conventional speaker identification system. Experiments demonstrate that the proposed method achieves significant performance improvements over previous missing feature techniques under a wide range of signal-to-noise ratios.
@InProceedings{6952536,
author = {X. Zhang and H. Zhang and G. Gao},
booktitle = {2014 22nd European Signal Processing Conference (EUSIPCO)},
title = {Missing feature reconstruction methods for robust speaker identification},
year = {2014},
pages = {1482-1486},
abstract = {In this study, we propose a reconstruction method to restore the degraded features for robust speaker identification. The proposed method is based on a hybrid generative model which consists of deep belief network (DBN) and restricted Boltzmann machine (RBM). Specifically, the noisy speech is firstly decomposed into time-frequency (T-F) representations. Then ideal binary mask (IBM) is computed to indicate each T-F point as reliable or unreliable. We reconstruct the unreliable ones by the proposed model iteratively. Finally, reconstructed feature is utilized to conventional speaker identification system. Experiments demonstrate that the proposed method achieves significant performance improvements over previous missing feature techniques under a wide range of signal-to-noise ratios.},
keywords = {signal reconstruction;signal representation;signal restoration;speaker recognition;time-frequency analysis;missing feature reconstruction methods;robust speaker identification system;deep belief network;hybrid generative model;DBN;restricted Boltzmann machine;RBM;noisy speech;time-frequency representations;T-F representations;ideal binary mask;IBM;T-F point;signal-to-noise ratios;Robustness;Abstracts;Computational modeling;Adaptation models;Data models;Production facilities;Smoothing methods;Robust speaker identification;Missing feature techniques;Restricted Boltzmann machine;Deep belief network},
issn = {2076-1465},
month = {Sep.},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925057.pdf},
}
Downloads: 0
{"_id":"rKSGNmfw7KkFuEnSm","bibbaseid":"zhang-zhang-gao-missingfeaturereconstructionmethodsforrobustspeakeridentification-2014","authorIDs":[],"author_short":["Zhang, X.","Zhang, H.","Gao, G."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["X."],"propositions":[],"lastnames":["Zhang"],"suffixes":[]},{"firstnames":["H."],"propositions":[],"lastnames":["Zhang"],"suffixes":[]},{"firstnames":["G."],"propositions":[],"lastnames":["Gao"],"suffixes":[]}],"booktitle":"2014 22nd European Signal Processing Conference (EUSIPCO)","title":"Missing feature reconstruction methods for robust speaker identification","year":"2014","pages":"1482-1486","abstract":"In this study, we propose a reconstruction method to restore the degraded features for robust speaker identification. The proposed method is based on a hybrid generative model which consists of deep belief network (DBN) and restricted Boltzmann machine (RBM). Specifically, the noisy speech is firstly decomposed into time-frequency (T-F) representations. Then ideal binary mask (IBM) is computed to indicate each T-F point as reliable or unreliable. We reconstruct the unreliable ones by the proposed model iteratively. Finally, reconstructed feature is utilized to conventional speaker identification system. Experiments demonstrate that the proposed method achieves significant performance improvements over previous missing feature techniques under a wide range of signal-to-noise ratios.","keywords":"signal reconstruction;signal representation;signal restoration;speaker recognition;time-frequency analysis;missing feature reconstruction methods;robust speaker identification system;deep belief network;hybrid generative model;DBN;restricted Boltzmann machine;RBM;noisy speech;time-frequency representations;T-F representations;ideal binary mask;IBM;T-F point;signal-to-noise ratios;Robustness;Abstracts;Computational modeling;Adaptation models;Data models;Production facilities;Smoothing methods;Robust speaker identification;Missing feature techniques;Restricted Boltzmann machine;Deep belief network","issn":"2076-1465","month":"Sep.","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925057.pdf","bibtex":"@InProceedings{6952536,\n author = {X. Zhang and H. Zhang and G. Gao},\n booktitle = {2014 22nd European Signal Processing Conference (EUSIPCO)},\n title = {Missing feature reconstruction methods for robust speaker identification},\n year = {2014},\n pages = {1482-1486},\n abstract = {In this study, we propose a reconstruction method to restore the degraded features for robust speaker identification. The proposed method is based on a hybrid generative model which consists of deep belief network (DBN) and restricted Boltzmann machine (RBM). Specifically, the noisy speech is firstly decomposed into time-frequency (T-F) representations. Then ideal binary mask (IBM) is computed to indicate each T-F point as reliable or unreliable. We reconstruct the unreliable ones by the proposed model iteratively. Finally, reconstructed feature is utilized to conventional speaker identification system. Experiments demonstrate that the proposed method achieves significant performance improvements over previous missing feature techniques under a wide range of signal-to-noise ratios.},\n keywords = {signal reconstruction;signal representation;signal restoration;speaker recognition;time-frequency analysis;missing feature reconstruction methods;robust speaker identification system;deep belief network;hybrid generative model;DBN;restricted Boltzmann machine;RBM;noisy speech;time-frequency representations;T-F representations;ideal binary mask;IBM;T-F point;signal-to-noise ratios;Robustness;Abstracts;Computational modeling;Adaptation models;Data models;Production facilities;Smoothing methods;Robust speaker identification;Missing feature techniques;Restricted Boltzmann machine;Deep belief network},\n issn = {2076-1465},\n month = {Sep.},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925057.pdf},\n}\n\n","author_short":["Zhang, X.","Zhang, H.","Gao, G."],"key":"6952536","id":"6952536","bibbaseid":"zhang-zhang-gao-missingfeaturereconstructionmethodsforrobustspeakeridentification-2014","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925057.pdf"},"keyword":["signal reconstruction;signal representation;signal restoration;speaker recognition;time-frequency analysis;missing feature reconstruction methods;robust speaker identification system;deep belief network;hybrid generative model;DBN;restricted Boltzmann machine;RBM;noisy speech;time-frequency representations;T-F representations;ideal binary mask;IBM;T-F point;signal-to-noise ratios;Robustness;Abstracts;Computational modeling;Adaptation models;Data models;Production facilities;Smoothing methods;Robust speaker identification;Missing feature techniques;Restricted Boltzmann machine;Deep belief network"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2014url.bib","creationDate":"2021-02-13T17:43:41.695Z","downloads":0,"keywords":["signal reconstruction;signal representation;signal restoration;speaker recognition;time-frequency analysis;missing feature reconstruction methods;robust speaker identification system;deep belief network;hybrid generative model;dbn;restricted boltzmann machine;rbm;noisy speech;time-frequency representations;t-f representations;ideal binary mask;ibm;t-f point;signal-to-noise ratios;robustness;abstracts;computational modeling;adaptation models;data models;production facilities;smoothing methods;robust speaker identification;missing feature techniques;restricted boltzmann machine;deep belief network"],"search_terms":["missing","feature","reconstruction","methods","robust","speaker","identification","zhang","zhang","gao"],"title":"Missing feature reconstruction methods for robust speaker identification","year":2014,"dataSources":["A2ezyFL6GG6na7bbs","oZFG3eQZPXnykPgnE"]}