rssfeed

WFSTs represent a means of pre-compiling the grammars used by ASR decoders. Such pre-compiled grammars tend to lead to both quicker decoding, because duplication of weights and paths is minimised, and simpler decoder implementations, because the composition process is offline. Pre-compiled grammars are quite large, sometimes limited only by the size of the memory on the host CPU.

Data-flow is a signal processing paradigm where modular components can be arranged in directed graphs to perform complex tasks. Data-flow lends itself to block calculations that take advantage of the hierarchical caches on modern CPUs. It also naturally modularises flow-based algorithms such that the components are easily re-used.

Context modelling for object detection.

On July 16 2010, Joel Pinto successfully defended his PhD thesis entitled "Multilayer Perceptron Based Hierarchical Acoustic Modeling for Automatic Speech Recognition".

Active participation of IM2 in a festival for children

Dr Daniel Gatica-Perez, (from the Idiap Research Institute, in Martigny, Switzerland): “People who are exploring ancient Maya culture – in order to try to find glyphs, they usually use catalogues and this is a classic catalogue. It’s like a telephone book. Instead of names what we have are essentially glyphs and this is a very large collection. And, essentially, if you don’t have computational means to find a glyph you don’t know or a glyph you are learning about, you probably have to parse all of these manually. And this can be time consuming.”

L'Idiap se rapproche toujours plus de l'EPFL. Grâce à une alliance, les chercheurs de l'institut martignerain auront accès aux titres académiques. Cela permettra notamment à l'Idiap, qui emploie aujourd'hui une centaine de scientifiques, de conserver ses meilleurs chercheurs. Le directeur adjoint, Jean-Albert Ferrez, est l’invité de L’Emission en Direct de Canal 9.

En 2010, Indiana Jones troquerait certainement son fouet contre un ordinateur ou un rayon laser. Depuis quelques années, grâce à un arsenal de nouvelles technologies, les archéologues ont avancé à grands pas. Dans les inventions récentes, il y a ce moteur de recherche capable de déchiffrer l’écriture maya.

The story follows the customer experience of a mobile wallet user. The level of security through biometric authentication increases with the amount of the transaction (no authentication for a soda, face authentication for a 40-Euros purchase, face and voice authentications for a cash withdrawal or a purchase above 100 Euros)

This group is open to everyone who is interested in the Idiap's life and activities.

He will be working with Daniel Gatica on "methods for activity representations from video".

Web-fying real life is the vision of the AMI consortium, which stands for Augmented Multiparty Interaction. The consortium is the force behind two highly successful European projects, AMI and AMIDA

Pippo is a PeopleBot , i.e. a mobile robot able to move indoors on flat floors. It is 112 cm high and it is equipped with a gripper, a table-sensing IR's, a precise pan-tilt camera, lasers and microphones.

Idiap just received the new prototype of a microphone array device which includes the improvements developed during last year project (see previous news below). No more need of bulky cables (see comparison picture) which are now replaced by a simple USB connexion, but also no more need to carry an external soundcard with you; all you need is your Mac and the microphone array.

Klewel in partnership with the Idiap Research Institute has conducted an AMIDA mini-project on automatic captioning of video lectures.

Laurent El Shafey has joined Idiap in January to work on Face Recognition.