Abstract
A fuzzy modelling approach to hidden Markov models (HMMs) for speech recognition is presented in this paper. State sequences are viewed as fuzzy sets and a timedependent fuzzy membership function is defined to represent the degree of belonging of an observation sequence to fuzzy state sequences at each time. An optimisation criterion for fuzzy HMMs is also presented. Parameter reestimation equations are derived from this criterion and used to train fuzzy HMMs for speech recognition. Fuzzy models are more effective than conventional models in isolated word recognition performed on the TI46 speech data corpus.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the Ninth Australian International Conference on Speech Science and Technology |
| Place of Publication | Melbourne, Australia |
| Publisher | Australian Speech Science and Technology Associatn |
| Pages | 473-478 |
| Number of pages | 6 |
| ISBN (Print) | 0 9581946 0 2 |
| Publication status | Published - 2002 |
| Event | Ninth Australian International Conference on Speech Science and Technology - Melbourne, Australia Duration: 3 Dec 2002 → 5 Dec 2002 |
Conference
| Conference | Ninth Australian International Conference on Speech Science and Technology |
|---|---|
| Country/Territory | Australia |
| City | Melbourne |
| Period | 3/12/02 → 5/12/02 |