Cluster-based adaptation using density forest for HMM phone recognition. Abou-Zleikha, M., Tan, Z., Christensen, M. G., & Jensen, S. H. In 2014 22nd European Signal Processing Conference (EUSIPCO), pages 2065-2069, Sep., 2014.
Paper abstract bibtex The dissimilarity between the training and test data in speech recognition systems is known to have a considerable effect on the recognition accuracy. To solve this problem, we use density forest to cluster the data and use maximum a posteriori (MAP) method to build a cluster-based adapted Gaussian mixture models (GMMs) in HMM speech recognition. Specifically, a set of bagged versions of the training data for each state in the HMM is generated, and each of these versions is used to generate one GMM and one tree in the density forest. Thereafter, an acoustic model forest is built by replacing the data of each leaf (cluster) in each tree with the corresponding GMM adapted by the leaf data using the MAP method. The results show that the proposed approach achieves 3:8% (absolute) lower phone error rate compared with the standard HMM/GMM and 0:8% (absolute) lower PER compared with bagged HMM/GMM.
@InProceedings{6952753,
author = {M. Abou-Zleikha and Z. Tan and M. G. Christensen and S. H. Jensen},
booktitle = {2014 22nd European Signal Processing Conference (EUSIPCO)},
title = {Cluster-based adaptation using density forest for HMM phone recognition},
year = {2014},
pages = {2065-2069},
abstract = {The dissimilarity between the training and test data in speech recognition systems is known to have a considerable effect on the recognition accuracy. To solve this problem, we use density forest to cluster the data and use maximum a posteriori (MAP) method to build a cluster-based adapted Gaussian mixture models (GMMs) in HMM speech recognition. Specifically, a set of bagged versions of the training data for each state in the HMM is generated, and each of these versions is used to generate one GMM and one tree in the density forest. Thereafter, an acoustic model forest is built by replacing the data of each leaf (cluster) in each tree with the corresponding GMM adapted by the leaf data using the MAP method. The results show that the proposed approach achieves 3:8% (absolute) lower phone error rate compared with the standard HMM/GMM and 0:8% (absolute) lower PER compared with bagged HMM/GMM.},
keywords = {Gaussian processes;hidden Markov models;maximum likelihood estimation;speech recognition;cluster-based adaptation;HMM phone recognition;speech recognition systems;density forest;maximum a posteriori method;MAP method;Gaussian mixture models;GMM;acoustic model forest;hidden Markov models;Hidden Markov models;Vegetation;Speech recognition;Data models;Speech;Acoustics;Adaptation models;ensemble acoustic modeling;density forest;cluster-based adaptation;HMM speech recognition},
issn = {2076-1465},
month = {Sep.},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925095.pdf},
}
Downloads: 0
{"_id":"mbueSNWA37QDhaQiz","bibbaseid":"abouzleikha-tan-christensen-jensen-clusterbasedadaptationusingdensityforestforhmmphonerecognition-2014","authorIDs":[],"author_short":["Abou-Zleikha, M.","Tan, Z.","Christensen, M. G.","Jensen, S. H."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["M."],"propositions":[],"lastnames":["Abou-Zleikha"],"suffixes":[]},{"firstnames":["Z."],"propositions":[],"lastnames":["Tan"],"suffixes":[]},{"firstnames":["M.","G."],"propositions":[],"lastnames":["Christensen"],"suffixes":[]},{"firstnames":["S.","H."],"propositions":[],"lastnames":["Jensen"],"suffixes":[]}],"booktitle":"2014 22nd European Signal Processing Conference (EUSIPCO)","title":"Cluster-based adaptation using density forest for HMM phone recognition","year":"2014","pages":"2065-2069","abstract":"The dissimilarity between the training and test data in speech recognition systems is known to have a considerable effect on the recognition accuracy. To solve this problem, we use density forest to cluster the data and use maximum a posteriori (MAP) method to build a cluster-based adapted Gaussian mixture models (GMMs) in HMM speech recognition. Specifically, a set of bagged versions of the training data for each state in the HMM is generated, and each of these versions is used to generate one GMM and one tree in the density forest. Thereafter, an acoustic model forest is built by replacing the data of each leaf (cluster) in each tree with the corresponding GMM adapted by the leaf data using the MAP method. The results show that the proposed approach achieves 3:8% (absolute) lower phone error rate compared with the standard HMM/GMM and 0:8% (absolute) lower PER compared with bagged HMM/GMM.","keywords":"Gaussian processes;hidden Markov models;maximum likelihood estimation;speech recognition;cluster-based adaptation;HMM phone recognition;speech recognition systems;density forest;maximum a posteriori method;MAP method;Gaussian mixture models;GMM;acoustic model forest;hidden Markov models;Hidden Markov models;Vegetation;Speech recognition;Data models;Speech;Acoustics;Adaptation models;ensemble acoustic modeling;density forest;cluster-based adaptation;HMM speech recognition","issn":"2076-1465","month":"Sep.","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925095.pdf","bibtex":"@InProceedings{6952753,\n author = {M. Abou-Zleikha and Z. Tan and M. G. Christensen and S. H. Jensen},\n booktitle = {2014 22nd European Signal Processing Conference (EUSIPCO)},\n title = {Cluster-based adaptation using density forest for HMM phone recognition},\n year = {2014},\n pages = {2065-2069},\n abstract = {The dissimilarity between the training and test data in speech recognition systems is known to have a considerable effect on the recognition accuracy. To solve this problem, we use density forest to cluster the data and use maximum a posteriori (MAP) method to build a cluster-based adapted Gaussian mixture models (GMMs) in HMM speech recognition. Specifically, a set of bagged versions of the training data for each state in the HMM is generated, and each of these versions is used to generate one GMM and one tree in the density forest. Thereafter, an acoustic model forest is built by replacing the data of each leaf (cluster) in each tree with the corresponding GMM adapted by the leaf data using the MAP method. The results show that the proposed approach achieves 3:8% (absolute) lower phone error rate compared with the standard HMM/GMM and 0:8% (absolute) lower PER compared with bagged HMM/GMM.},\n keywords = {Gaussian processes;hidden Markov models;maximum likelihood estimation;speech recognition;cluster-based adaptation;HMM phone recognition;speech recognition systems;density forest;maximum a posteriori method;MAP method;Gaussian mixture models;GMM;acoustic model forest;hidden Markov models;Hidden Markov models;Vegetation;Speech recognition;Data models;Speech;Acoustics;Adaptation models;ensemble acoustic modeling;density forest;cluster-based adaptation;HMM speech recognition},\n issn = {2076-1465},\n month = {Sep.},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925095.pdf},\n}\n\n","author_short":["Abou-Zleikha, M.","Tan, Z.","Christensen, M. G.","Jensen, S. H."],"key":"6952753","id":"6952753","bibbaseid":"abouzleikha-tan-christensen-jensen-clusterbasedadaptationusingdensityforestforhmmphonerecognition-2014","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2014/html/papers/1569925095.pdf"},"keyword":["Gaussian processes;hidden Markov models;maximum likelihood estimation;speech recognition;cluster-based adaptation;HMM phone recognition;speech recognition systems;density forest;maximum a posteriori method;MAP method;Gaussian mixture models;GMM;acoustic model forest;hidden Markov models;Hidden Markov models;Vegetation;Speech recognition;Data models;Speech;Acoustics;Adaptation models;ensemble acoustic modeling;density forest;cluster-based adaptation;HMM speech recognition"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2014url.bib","creationDate":"2021-02-13T17:43:41.753Z","downloads":0,"keywords":["gaussian processes;hidden markov models;maximum likelihood estimation;speech recognition;cluster-based adaptation;hmm phone recognition;speech recognition systems;density forest;maximum a posteriori method;map method;gaussian mixture models;gmm;acoustic model forest;hidden markov models;hidden markov models;vegetation;speech recognition;data models;speech;acoustics;adaptation models;ensemble acoustic modeling;density forest;cluster-based adaptation;hmm speech recognition"],"search_terms":["cluster","based","adaptation","using","density","forest","hmm","phone","recognition","abou-zleikha","tan","christensen","jensen"],"title":"Cluster-based adaptation using density forest for HMM phone recognition","year":2014,"dataSources":["A2ezyFL6GG6na7bbs","oZFG3eQZPXnykPgnE"]}