Compact Acoustic Models for Embedded Speech Recognition

Abstract : Recommended by Joe Picone Speech recognition applications are known to require a significant amount of resources. However, embedded speech recognition only authorizes few KB of memory, few MIPS, and small amount of training data. In order to fit the resource constraints of embedded applications, an approach based on a semicontinuous HMM system using state-independent acoustic modelling is proposed. A transformation is computed and applied to the global model in order to obtain each HMM state-dependent probability density functions, authorizing to store only the transformation parameters. This approach is evaluated on two tasks: digit and voice-command recognition. A fast adaptation technique of acoustic models is also proposed. In order to significantly reduce computational costs, the adaptation is performed only on the global model (using related speaker recognition adaptation techniques) with no need for state-dependent data. The whole approach results in a relative gain of more than 20% compared to a basic HMM-based system fitting the constraints.
Document type :
Journal articles
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01318188
Contributor : Bibliothèque Universitaire Déposants Hal-Avignon <>
Submitted on : Thursday, May 19, 2016 - 12:35:34 PM
Last modification on : Saturday, June 15, 2019 - 12:24:17 PM

Links full text

Identifiers

Collections

Citation

Christophe Lévy, Georges Linarès, Jean-François Bonastre. Compact Acoustic Models for Embedded Speech Recognition. EURASIP Journal on Audio, Speech, and Music Processing, SpringerOpen, 2009, ⟨10.1155/2009/806186⟩. ⟨hal-01318188⟩

Share

Metrics

Record views

537