Difference between revisions of "Skills"

From LRDE

Line 15: Line 15:
 
__NOTOC__
 
__NOTOC__
   
== <span style="color:#000080">'''Theme 1: Image and pattern recognition'''</span> ==
+
== <span style="color:#000080">'''Theme 1: Image processing and pattern recognition'''</span> ==
   
 
=== '''Image processing''' ===
 
=== '''Image processing''' ===
Line 24: Line 24:
 
=== '''Speaker ID and signal processing''' ===
 
=== '''Speaker ID and signal processing''' ===
 
In the field of speaker recognition, the LRDE works on issues related to channel effects, the variability of the speaker and, more generally, to mismatch conditions between training and test speech utterances. The [[Speaker ID]]-team proposes low-dimensional representations of speaker identities.
 
In the field of speaker recognition, the LRDE works on issues related to channel effects, the variability of the speaker and, more generally, to mismatch conditions between training and test speech utterances. The [[Speaker ID]]-team proposes low-dimensional representations of speaker identities.
 
 
   
 
== <span style="color:#000080">'''Theme 2: Automata and verification''' </span> ==
 
== <span style="color:#000080">'''Theme 2: Automata and verification''' </span> ==

Revision as of 13:27, 30 October 2013

This page quickly presents the areas of expertise of the LRDE. For more detailed information, please go to the projects page.


The activities of LRDE cover two main areas of research:

  • « Image and pattern recognition » and
  • « Automata and verification »

with a transverse research axis « Performance and genericity ».

Each main area, or theme, is divided into two fields.



Theme 1: Image processing and pattern recognition

Image processing

Image processing at LRDE consists mainly of writing morphological algorithms. The principal fields of application are document image analysis and medical image processing. The state of the art of these elements as well as all the results of the Image-team's research have been integrated in Olena, a high performance generic image processing library.


Speaker ID and signal processing

In the field of speaker recognition, the LRDE works on issues related to channel effects, the variability of the speaker and, more generally, to mismatch conditions between training and test speech utterances. The Speaker ID-team proposes low-dimensional representations of speaker identities.

Theme 2: Automata and verification

Multiple finite-state machines

LRDE's Automata-team develops Vaucanson, a platform for the manipulation of multiple finite-state machines in order to provide genericity, efficiency, respect of the theory and multipurpose tools.

Formal verification

The second aspect of the team's research concerns automata on infinite words (ω-words) used for formal verification or model checking. The Spot library has been designed for this.


Transverse research axis

A common underlying problem of all the projects LRDE members work on is how to reconcile two aspects usually contradictory: performance and genericity.