GMM and HMM training by aggregated EM algorithm with increased ensemble sizes for robust parameter estimation

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

Volumn , Issue , 2008, Pages 4405-4408

  Shinozaki, Takahiro ^a,b   Kawahara, Tatsuya ^a  

^a KYOTO UNIVERSITY   (Japan)

^b TOKYO INSTITUTE OF TECHNOLOGY   (Japan)

Author keywords

Bagging; Ensemble training; Expectation maximization algorithm; Hidden Markov model; Sufficient statistics

Indexed keywords

ACOUSTICS; AGGLOMERATION; ALGORITHMS; PARAMETER ESTIMATION; SIGNAL PROCESSING; SILVER; SPEECH; SPEECH RECOGNITION;

BAGGING; ENSEMBLE TRAINING; EXPECTATION MAXIMIZATION ALGORITHM; HIDDEN MARKOV MODEL; SUFFICIENT STATISTICS;

MATHEMATICAL MODELS;

EID: 51449090592     PISSN: 15206149     EISSN: None     Source Type: Conference Proceeding    
DOI: 10.1109/ICASSP.2008.4518632     Document Type: Conference Paper

References (5)

1. T. Shinozaki and M. Ostendorf, "Cross-validation EM training for robust parameter estimation," in ICASSP, Hawaii, 2007, vol. IV, pp. 437-440.
2. T. Shinozaki and M. Ostendorf, "Cross-validation and aggregated EM training for robust parameter estimation," Computer speech and language, accepted.
3. S. Young et al., The HTK Book, Cambridge University Engineering Department, 2005.
4. T. Kawahara, H. Nanjo, T. Shinozaki, and S. Furui, "Benchmark test for speech recognition using the Corpus of Spontaneous Japanese," in Proc. SSPR2003, 2003, pp. 135-138.
5. A. Lee, T. Kawahara, and S. Doshita, "An efficient two-pass search algorithm using word trellis index," in Proc. ICSLP, 1998, pp. 1831-1834.