#### DMCA

## Scaled random trajectory segment models (1998)

Venue: | Computer Speech and Language |

Citations: | 2 - 1 self |

### Citations

609 | Maximum likelihood estimation from incomplete data via the EM algorithm (with Discussion). - Dempster, Laird, et al. - 1977 |

124 |
The segmential K-means algorithm for estimating parameters of hidden Markov models.
- Juang, Rabiner
- 1990
(Show Context)
Citation Context ...he algorithm attempts to bring the objective function to a maximum. The other estimation scheme is an approximation of the Baum-Welch method, based on an extension of the segmental K-means algorithm (=-=Juang and Rabiner, 1990-=-). Each iteration is composed of two steps. In thesrst stage, a Viterbi decoding algorithm is applied to obtain the most likely state sequence (using the current values of the estimated parameters). O... |

63 |
Speech recognition using hidden Markov models with polynomial regression functions as non-stationary states.
- Deng, Aksmanovic, et al.
- 1994
(Show Context)
Citation Context ... acoustic models we implemented for evaluation were: 1. Standard Gaussian HMM. 2. Static RTSM (Russell, 1993). 3. Scaled static RTSM (presented in section 3). 4. Linear mean trajectory segment model (=-=Deng et al., 1994-=-). 5. Linear RTSM (Holmes and Russell, 1995). 6. Scaled linear RTSM (presented in section 4). Several alternatives for model topologies were employed in order to analyze the balancing problem in dier... |

48 | Genones, generalized mixture tying in continuous hidden Markov model-based speech recognizers,” - Digalakis, Monaco, et al. - 1996 |

45 | Segmental hidden Markov models," in
- Gales, Young
- 1995
(Show Context)
Citation Context ...f the static RTSM, the probability expressions are formulated in terms of the shift of the trajectory from the model mean, rather than the location of the trajectory itself as in (Russell, 1993) and (=-=Gales and Young, 1993-=-). We are adopting this convention because it explicitly re ects the linear behavior of the model (equation 1). Two methods for computing the likelihood score have been proposed. Thesrst one (Russell,... |

28 |
Hidden Markove models with templates as non-stationary states: an application to speech recognition
- Ghitza, Sondhi
- 1993
(Show Context)
Citation Context ...of several segments). This modeling approach is opposed to the HMM paradigm, that utilizes frame-based modeling. Various segment models have been proposed, e.g. (Deng, Aksmanovic, Sun and Wu, 1994), (=-=Ghitza and Sondhi, 1993-=-), (Digalakis, Rohlicek and Ostendorf, 1993) and (Goldberger, Burshtein and Franco, 1997). A comprehensive survey on the subject can be found in (Ostendorf, Digalakis and Kimball, 1996). A number of s... |

27 | Segment-based stochastic models of spectral dynamics for continuous speech recognition,” - Digalakis - 1992 |

19 | A dynamical system approach to continuous speech recognition - Digalakis, Ostendorf, et al. - 1992 |

18 | From HMM's to segment models: A unied view of stochastic modeling for speech recognition. - Ostendorf, Digalakis, et al. - 1996 |

15 |
Speech recognition using a linear dynamic segmental HMM
- Holmes, Russell
- 1995
(Show Context)
Citation Context ...evaluation were: 1. Standard Gaussian HMM. 2. Static RTSM (Russell, 1993). 3. Scaled static RTSM (presented in section 3). 4. Linear mean trajectory segment model (Deng et al., 1994). 5. Linear RTSM (=-=Holmes and Russell, 1995-=-). 6. Scaled linear RTSM (presented in section 4). Several alternatives for model topologies were employed in order to analyze the balancing problem in dierent situations. Thesrst analyzed topology a... |

12 | Modeling speech variability with segmental HMMs - Holmes, Russell - 1996 |

9 | Segmental modeling using a continuous mixture of nonparametric models - Goldberger, Burshtein, et al. - 1999 |

8 | Linear dynamic segmental HMMS : Variability representation and training procedure - Holmes, Russell - 1997 |

1 | Speaker independent phonetic classi using hidden Markov models with state conditioned mixtures of trend functions - Deng, Aksmanovic - 1997 |