!!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!! !!!
-------------------------------------------------------------------------------
                                       
                                       
                   B O S T O N   U N I V E R S I T Y
                                       
                      Computer Science Department
                                       
                          C O L L O Q U I U M

                      Motion Pattern Analysis and 
                Vision-Based Human Computer Interactions


			    Ming-Hsuan Yang
			  Computer Vision Lab
			   Beckman Institute
		University of Illinois at Urbana-Champaign

 			Wednesday, February 23 
                                      
                               11:00 AM
                       (Coffee served at 10:45AM)

                                       
                         Seminar Room / MCS 135





    Vision-based Human-Computer Interactions have become an active research
    area and produced exciting applications in recent years. Such systems
    usually require not only advanced vision techniques to extract low level
    information but also an extensive use of learning methods that can
    generalize from previous experiences to new situations.  

    This talk presents studies on the combination of low level visual
    processing with sophisticated learning methods for of visual recognition
    tasks that are at the center of Vision-based Human-Computer intelligent
    Interactions. I will present a vision algorithm for extracting and
    classifying two-dimensional motion in an image sequence based on motion
    trajectories.  Motion patterns are learned from the extracted
    trajectories using a time-delay neural network.  We exhibit the advantages 
    of our method by evaluating it on 40 hand gestures of American Sign
    Language.  Experimental results show that motion patterns in hand gestures 
    can be extracted recognized with high recognition rate using motion 
    trajectories.  I will also briefly discuss works on novel learning 
    algorithms, face detection, object recognition and vision-based interfaces 
    if time permits.


    Host:  Stan Sclaroff (sclaroff@cs.bu.edu)

-------------------------------------------------------------------------------
For colloquium info, including directions, see http://cs-www.bu.edu/colloquium
-------------------------------------------------------------------------------