Unsupervised Stream-Weights Computation in Classification and Recognition Tasks - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Journal Articles IEEE Transactions on Audio, Speech and Language Processing Year : 2009

Unsupervised Stream-Weights Computation in Classification and Recognition Tasks

Abstract

In this paper, we provide theoretical results on the problem of optimal stream weight selection for the multi-stream classi- fication problem. It is shown, that in the presence of estimation or modeling errors using stream weights can decrease the total classification error. Stream weight estimates are computed for various conditions. Then we turn our attention to the problem of unsupervised stream weights computation. Based on the theoretical results we propose to use models and “anti-models” (class- specific background models) to estimate stream weights. A non-linear function of the ratio of the inter- to intra-class distance is used for stream weight estimation. The proposed unsupervised stream weight estimation algorithm is evaluated on both artificial data and on the problem of audio-visual speech classification. Finally the proposed algorithm is extended to the problem of audio- visual speech recognition. It is shown that the proposed algorithms achieve results comparable to the supervised minimum-error training approach under most testing conditions.
Fichier principal
Vignette du fichier
Final-TrSAL09.pdf (754.16 Ko) Télécharger le fichier
Origin : Publisher files allowed on an open archive
Loading...

Dates and versions

inria-00438846 , version 1 (04-12-2009)

Identifiers

  • HAL Id : inria-00438846 , version 1

Cite

Eduardo Sanchez-Soto, Alexandros Potamianos, Khalid Daoudi. Unsupervised Stream-Weights Computation in Classification and Recognition Tasks. IEEE Transactions on Audio, Speech and Language Processing, 2009, 17 (3), pp.436--445. ⟨inria-00438846⟩
177 View
346 Download

Share

Gmail Facebook X LinkedIn More