Invited Talk by Prof. Hironobu Fujiyoshi
People Image Analysis - Detecting Humans and Visualizing Human Motions -
Abstract:
An invisible environmental robot is a system that recognizes ever-changing human states based on input provided by a group of sensors such as cameras situated in an environment. An important role of such systems is to help create a comfortable space for human users. The quality of user actions can be enhanced by sensing the situations of humans including their movements, trajectory, posture, line of sight and facial expressions, and supporting the recognition of human behavior intentions. In the sensing technology, people image analysis (PIA) is essential to recognize the ever-changing human situations in real time. The key technologies for PIA are: detecting humans in moving images, tracking their movements, detecting faces, tracking the location of face parts, and understanding motions. Of these, we introduce the techniques used for the detection of humans and the visualization of human motions used to connect this sensing information with activity recognition.
Biography:
Dr. Fujiyoshi is a professor of Computer Science, Chubu University, Japan, where his research interests are computer vision, video understanding and pattern recognition based on machine learning. He received his Ph.D. in Electrical Engineering from Chubu University in 1977, for his research on fingerprint verification using spectrum analysis, which was introduced into a manufactured device by a security company. From 1997 to 2000 he was a post-doctoral fellow at the Robotics Institute of Carnegie Mellon University, USA, working on the DARPA Video Surveillance and Monitoring effort and the humanoid vision project for the HONDA Humanoid Robot. From 2005 to 2006, he also worked on the people image analysis project as a visiting researcher at Carnegie Mellon University.