• E-mail
  
• Home Page
  

Home
People
	Department Contacts
	Faculty
	images
	research
	contact
	Graduate Students
	Undergraduate Students
	Research Staff
	Visitors
	Admin. & Technical Staff
Academics
Research
Resources

Peter J. Ramadge

My research interests are in the areas of signal processing and systems. I have worked on problems of adaptive signal processing, adaptive control, discrete-event systems, optimization of queueing systems, and on video analysis, annotation, and search.

Recent research on digital video libraries has centered on developing efficient algorithms for searching compressed digital video and audio signals. For example, we have developed fast algorithms for estimating camera motion directly from MPEG compressed video. When the camera motion satisfies certain constraints, then estimates of camera motion from the MPEG file are surprisingly accurate and can be used in suitable contexts as a useful indication of video content. We have also developed efficient algorithms for measuring the similarity of two video clips and for searching compressed video to find segments that are similar to a given example clip. This line of investigation is aimed at providing useful tools for searching a digital video/audio library to find content of interest.

A related area of research is the registration of images and video from multiple cameras. Once the frames of the videos have been adequately registered, they can be processed to create new views either for presentation or analysis purposes. Image registration is an important step, for example, in tomography and medical imaging. Registered video frames can be used to form a mosaic. This gives a compact representation of the video as well as a new way to visualize the data.

Another application is the creation of the view from a new virtual camera located in a restricted region of space relative to the actual cameras. This requires the correspondence between the frames and the virtual view to be accurately determined so that the virtual view can be rendered from the given data. Another line of recent research is concerned with the stability and performance of switched adaptive algorithms for predication and control. In this research problem one has available a set of prediction algorithms for a data sequence, and one wants to design a switching rule for switching among these predictors so that one eventually finds a "good" predictor for the data. This rule can form the basis of a compression scheme, or it can be used to make subsequent decisions about the generation of the data, for example, in control applications. When used as an adaptive control structure, a switching rule is a form of logical supervision, and the entire system becomes a hybrid (continuous and discrete) dynamical system. Of particular interest is the stability and asymptotic performance of these adaptive schemes.

I am the recipient of several teaching awards, an IBM faculty development award, and an IEEE best paper award. I am a fellow of the IEEE and a member of SIAM.