Signing Avatars

In the Signing Avatar Project we aim at combining avatar technology and linguistics to bridge the communication gap between the deaf and hearing people. Our goal is to develop an expressive digital French-to-LSF translator. We use Motion Capture Technology, including facial, hand and body data to synthesize more natural movements. We use concatenative synthesis for building new phrases in LSF from pre-recorded motion.

Related Publications

General

Hands

Facial Expressions

Motion Analysis and Synthesis from Captured Data

Motion descriptors can be computed from motion capture data. They range from low-level descriptors such as positions, velocities, accelerations to high-level descriptors such as the Effort descriptors of the Laban Movement Analysis (LMA). Once computed, those descriptors can be used for various applications: synthesis of motion, physically based 3D simulation, physically based sound synthesis, or motion classification and retrieval from databases.

Related Publications

Modeling

Models to be animated require polygonal meshes to be well-aligned with the silhouette of the character, are often symmetrical and mainly made of quads. Most automatic modeling techniques do not take these requirements into account. Manually created models are slow to produce, require a lot of do-undo work and a skilled artist. We thus aim at automatically producing 3D polygonal models from 2D drawings.

Related Publications

Character Animation

Skinning, also called Linear Blend Skinning (LBS) or Skeleton Subspace Deformations (SSD) is the standard technique for character animation to date. The character is composed of an internal skeleton and a polygonal mesh tied together through a rigging process. Although very efficient, this technique suffers several drawbacks including the well-known collapsing joint and candy wrapper effects. In addition, due to the nature of the algorithm effect itself it is not directly possible to simulate muscle bulging, dynamics or dynamic wrinkles of skin or clothes. Our aim is to enhance existing real-time character animations based on the skinning techniques with muscles, dynamics and wrinkles to make them visually more appealing.

Related Publications

Muscles

Dynamics

Wrinkles

Natural Phenomena

Characters necessarily evolve in an environment and interact with objects. For example, fish swimming in the sea interact with anemones in different ways depending on the type of fish. The anemones themselves sway in the water in a similar way that trees sway in the wind and react to the fluid surrounding them. Fire burns things such as paper that reacts in crumpling when it looses water and propagates heat as it burns.

Related Publications

Sea Anemones and Seaweeds

Paper Burning and Crumpling

Perceptually based Algorithms

More accurate does not mean perceived more realistic. The progress in character animation and rendering that has been made in the last twenty years has brought us to the uncanny valley. To climb out of there, the current trend is to include perception into the rending/animation algorithms. To derive new perceptually based algorithms, perceptual studies are conducted to find out what effects are perceived by the human eye/brain, and in which conditions.

Perceptual studies are also developed to analyse expressiveness, meaning and style of human motion, and to classify motion captured data.

Related Publications

Art, Color Science & Computational Aesthetics

Computational Art aims at creating images using simulation algorithms or as a non-photorealistic rendering of an input image. Computational String Art is an example of non-photorealistic rendering of an input black and white photograph. Creating realistic images is one thing, creating aesthetic ones is another. Two of the important factors of aesthetics are shape and color and it is possible to derive objective metrics (aesthetics is not entirely subjective, altough it is common belief). There are thus measures of color aesthetics that have been incorporated into Color Order Systems such as the Munsell COS or Coloroid. Two questions we've adressed are how to produce and classify fluorescent colors (that change depending on the illuminant) and how to use COS to create aesthetically pleasing images ?

Related Publications