A new method for the automatic derivation of HMM topologies is presented. At first, the speech signal is segmented into acoustical units by using an Ergodic HMM. Then, a lattice structure is built from all the observed pronunciations. The lattice is thus pruned according to an information-theoretic criterion, aiming to preserve only the more characteristic event sequences. A circuit-free HMM topology is finally built after a proper state number assignment. The method naturally permits the sharing of phonetically-motivated observation densities within different HMM and states. Results for a speaker-independent recognition experiment are given.
Automatic derivation of HMM alternative pronunciation network topologies / Falaschi, Alessandro; M., Pucci. - STAMPA. - (1991), pp. 671-674. (Intervento presentato al convegno EUROSPEECH '91 Second European Conference on Speech Communication and Technology tenutosi a Genova, Italy nel September 24-26, 1991).
Automatic derivation of HMM alternative pronunciation network topologies
FALASCHI, Alessandro;
1991
Abstract
A new method for the automatic derivation of HMM topologies is presented. At first, the speech signal is segmented into acoustical units by using an Ergodic HMM. Then, a lattice structure is built from all the observed pronunciations. The lattice is thus pruned according to an information-theoretic criterion, aiming to preserve only the more characteristic event sequences. A circuit-free HMM topology is finally built after a proper state number assignment. The method naturally permits the sharing of phonetically-motivated observation densities within different HMM and states. Results for a speaker-independent recognition experiment are given.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
