Active Research,
|
The study of motion has been the connecting theme of my research in the past few years. In graphics I have used nonlinear dynamic systems to control and efficiently and simulate large number of autonomous agents. In vision I have been working on tracking: the recovery of motion.
For years computer vision researchers have been working on the problem of tracking objects and features across video footage. Recently, tracking of human faces has acquired an even more active role in the field, being a necessary step for higher level applications such as surveillance, recognition, and analysis of behavior. It is hard to develop robust computer vision techniques that require minimal, or none, human intervention for a large set of possible inputs.
There are two main competing approaches for object tracking: tracking 2D regions of the image, and tracking full 3D representations of the face. Each one of these approaches actually solves a different problem, and presents its own set of advantages and limitations.
Tracking of 2D regions is easier to implement, it makes fewer assumptions about the camera, and it is usually fast. Unfortunately, it only looks for the 2D region where the face is in the the images. Full 3D tracking, on the other hand, estimates orientation and translation of the face in the 3D world. Additionally, the extraction of deformations (face expressions) can be done in a more robust and natural way. The catch is that 3D tracking is much more complicated to achieve, it is computationally expensive, and less robust. Additionally, it usually requires 3D model of the face and a model of the camera.
In our group, we have several projects where we are interested in the behavior of the facial expressions over time. We have developed a technique for robust 3D tracking of deformable faces. We use different low-level, well known, computer vision algorithms to extract 2D information from the video sequence. Then, with a new technique we have been developing, we convert them to a meaningful distribution in the multidimensional deformation space. To achieve robust tracking we combine the information obtained from these different sources with a maximum likelihood estimator to do an optimal data fusion.
We track the orientation and translation of the face, as well as the the parameters that describe the movement of eyebrows, mouth, etc. These parameters can then be directly used by algorithms that do analysis of the movement.
It is very hard to make agents and computer procedures intelligent (it is after all a whole field by itself!), but in many areas the restriction and conditioning of what "intelligent" means can turn it in a easier and more tractable problem.
In many applications of Computer Graphics, such as virtual reality, games, and simulations, the computer has to control agents that will coexist, and perhaps even interact, with user controlled agents. Although for humans it is natural to be aware the surrounding environment and react accordingly, it is a whole different story on the computer side.
In our work we automate reactive behaviors for a computer guided agent. Tasks such as collision avoidance of moving obstacles and reaching an also moving target are achieved in a parameterized way, allowing each agent to be unique.
Soon there will be a dedicated page with more information and demos.
In the past years wavelets and filter banks have established themselves as a very powerful set of tools for processing, analysis and manipulation of the most various types of signals.
In Computer Graphics wavelets are an important tool for several different applications, such as surface modeling and representation, texture analysis and synthesis, radiosity, and rendering.
Check our book for a little bit more about it.
There are many interesting and challenging problems related to sound and speech manipulation. In the past (look at my MSc thesis) I've studied the characteristics and properties of transformations of audio signals, using both time and frequency representations. One nice result is the use of the Local Cosine Transform to accomplish time warping of audio signals avoiding large frequency distortions.