30.01.2013 Views

TSI report for the period 2005-2009 - Département Traitement du ...

TSI report for the period 2005-2009 - Département Traitement du ...

TSI report for the period 2005-2009 - Département Traitement du ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

11.2. Main Results 11. Multimedia (MM)<br />

with F. Tupin et L. Denis ENSML). A new grant on this subject (funded by DGA/REI) has been<br />

accepted and should start soon (in collaboration with J-F, Aujol (CMLA) and J-M. Nicolas). M.<br />

Sigelle has also been working in collaboration with W. Pieczinsky (Télécom SudParis), F. Tupin<br />

and D. Benboudjema on triplet Markov Random Fields AIMED TO texture analysis and indexing<br />

in <strong>the</strong> framework of <strong>the</strong> Info@Magic project.<br />

M. Sigelle started a collaboration with I. Jermyn (INRIA ARIANA) and S. Perreau (UNISA<br />

Adelaide Australia) on <strong>the</strong> topics of (discrete) diffusion processes, which can be applied both to<br />

modelling of traffic routing in ad hoc networks and to image restoration [2194, 2360].<br />

The studies of C. Faure on documents and images emphasized <strong>the</strong> role of communication and<br />

<strong>the</strong> visual modality. Digital and digitised documents are processed to facilitate in<strong>for</strong>mation access.<br />

Layout and logical structures are automaticaly detected in document images or in semi-structured<br />

digital documents. Applications were developped <strong>for</strong> <strong>the</strong> RNTL project InfRadio <strong>for</strong> which web<br />

documents were adapted to be read and activated on <strong>the</strong> small screens of mobile devices [2224].<br />

More recently, document image analysis was per<strong>for</strong>med <strong>for</strong> <strong>the</strong> digital library medic@ to assist<br />

<strong>the</strong> archivists in indexing and storing historical medical documents. New methods were proposed<br />

to structure <strong>the</strong> images of <strong>the</strong> pages and to extract relevant components such as <strong>the</strong> figure and<br />

caption pairs [2096, 2094, 2095]. To cope with ancient fonts difficult to recognise by OCR, word<br />

spotting methods were proposed to search <strong>for</strong> word-images similar to query words [2231, 2128,<br />

2129]. These works <strong>for</strong> medic@ are made in collaboration with <strong>the</strong> LIPADE (Univ. Paris V). In<br />

GEOservice, a joint project between several research teams of <strong>the</strong> Institut Télécom (C. Faure<br />

was prime), <strong>the</strong> visual modality was involved in a web service. Images were combined with text<br />

to provide multimodal egocentric instructions <strong>for</strong> guiding a mobile user in a building. As a natural<br />

complement of <strong>the</strong> visual modality, <strong>the</strong> gestural modality was studied in <strong>the</strong> context of humancomputer<br />

interaction where <strong>the</strong> users drew or wrote to communicate [2093, 2223, 2260, 2234,<br />

2086].<br />

11.2.4 Audio-visual Identity/Imposture and Virtual Worlds<br />

Faculty G. Chollet, C. Pelachaud, M. Sigelle, M. Charbit<br />

Main events G. Chollet and C. Pelachaud, general co-chairs of IVA’07; C. Pelachaud and T.<br />

Boubekeur, co-editor special issue on Facial Modeling, IEEE Computer Graphics and Applications,<br />

to appear in 2010; C. Pelachaud co-organizer of a Workshop held in conjunction<br />

with AAMAS <strong>2009</strong>; she is since 2007 secretary of <strong>the</strong> Humaine association on emotion;<br />

she is part of <strong>the</strong> selection committee of ANR CONTINT (since 2008), ANR Blanc CSD9<br />

Sciences Humaines et sociales (in <strong>2009</strong>).<br />

Projects IST NoE BIOSECURE (2004-2007), IV2 TechnoVision (2006-2007), IST SECURE-<br />

PHONE (<strong>2005</strong>-2007), IST NoE KSpace (<strong>2005</strong>-2008), INFOM@GIC (Cap Digital) (2006-<br />

<strong>2009</strong>), ANR KIVAOU (2008-2010), ANR MYBlog3D (2006-2010), CompanionAble: IP de<br />

IST (2008-2012), ANR blanc OUISPER (2006-<strong>2009</strong>), IST IP-CALLAS (2006-2010), IST<br />

STREP-SEMAINE (2008-2011), IST NoE-SSPNet (<strong>2009</strong>-2013), COST Action 2102 (2006-<br />

2010), ANR CECIL (<strong>2009</strong>-2011), ANR GV-Lex (<strong>2009</strong>-2011), ANR IMMEMO (<strong>2009</strong>-2011)<br />

Two main directions of investigation are present in this <strong>the</strong>me:<br />

Biometry and Speech/Face Syn<strong>the</strong>sis/Recognition/Verification<br />

The speech group was created in 1983 when Gérard Chollet joigned Télécom-ParisTech (called<br />

ENST at <strong>the</strong> time). The focus was centered on coding, syn<strong>the</strong>sis and recognition. In <strong>the</strong> 1990,<br />

speaker verification was added, followed by language identification five years ago. At that time,<br />

audio-visual speech and speaker recognition became a topic of interest. The Biosecure network<br />

of excellence was an opportunity to promote open-source software <strong>for</strong> major biometric modalities<br />

(face, voice, audio-visual speaker, signature, iris, hand shape...) This led to <strong>the</strong> publication of<br />

206

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!