ISCA Archive Eurospeech 1997
ISCA Archive Eurospeech 1997

Bottom-up and top-down state clustering for robust acoustic modeling

Cristina Chesta, Pietro Laface, Franco Ravera

In this paper we describe our experience with bottom- up and top- down state clustering techniques for the definition and training of robust acoustic-phonetic units. Using as a test-bed a speaker- independent telephone- speech isolated word recognition task with a vocabulary including 475 city names, we show that similar performances are obtained by tying the HMM states both with an agglomerative or a decision-tree clustering approach. Moreover, better results are obtained by a priori selecting the set of states that can be clustered, rather than relying solely on their acoustical similarity. In the bottom-up approach a stopping criterion for the furthest neighbor clustering procedure is proposed that does not require a threshold. In the top-down approach we show that a careful selected impurity function allows lookahead search to outperforms the classical decision tree growing algorithm.


doi: 10.21437/Eurospeech.1997-9

Cite as: Chesta, C., Laface, P., Ravera, F. (1997) Bottom-up and top-down state clustering for robust acoustic modeling. Proc. 5th European Conference on Speech Communication and Technology (Eurospeech 1997), 11-14, doi: 10.21437/Eurospeech.1997-9

@inproceedings{chesta97_eurospeech,
  author={Cristina Chesta and Pietro Laface and Franco Ravera},
  title={{Bottom-up and top-down state clustering for robust acoustic modeling}},
  year=1997,
  booktitle={Proc. 5th European Conference on Speech Communication and Technology (Eurospeech 1997)},
  pages={11--14},
  doi={10.21437/Eurospeech.1997-9},
  issn={1018-4074}
}