My name is Huadong Wu (sounds like wa-dong wu). I am a new Ph.D. in Robotics from the Robotics Institute, School of Computer Science, Carnegie Mellon University. My research interest is in intelligent sensing, which includes signal processing with distributed sensors, pattern recognition, and sensor fusion (or data fusion, information fusion). My advisor is Professor Mel Siegel.
I am now working in medical image processing research as a postdoctoral associate, in the the Department of Radiology, University of Pittsburgh Medical Center (UPMC). But I may still need to look for a permenant full-time job. Please check my resume for a quick review of my qualifications.
Let me give you an overview of my background and experience. I graduated from Shanghai Jiao Tong University with a B.S. degree in Precision Instruments (a double major in Mechanical Engineering and Electrical Engineering) in 1984. I completed a M.S. degree in Inertial Navigation (signal processing and system control in inertial navigation applications) in 1987. Then I worked for about 8.5 years as an engineer in the Robot & Automation Research Institute, Beijing, China. My major work was in mechanical design and mechanical-electrical system control. Most of this work was related to research and development of industrial robots.
During my internship in the Motorola Research Labs in Schaumburg, Illinois, in the summer of 2000, I became very interested in the sensor fusion challenges in context-aware computing research. I directed my Ph.D. dissertation research into this area. My dissertation topic is Sensor Data Fusion for Context-Aware Computing Using Dempster-Shafer Theory.
"Context-aware computing" is to make computers understand human situational "context" information and respond accordingly. To sense the context needs many sensors to collaborate, and because most context-aware computing applications are for direct human-computer interactions, the first challenge is how to fuse the sensors' objective observations with human's subjective opinions -- as human interventions are often desired. In addition, because the sensor set in a context-aware computing system is typically highly dynamic in configuration, traditional sensor fusion methods cannot overcome the difficulties that the sensors' observational statistics data are not enough and their joint probability distribution is usually unavailable.
The Dempster-Shafer Theory of Evidence reasoning framework is introduced to meet these challenges. In practice, to get around the difficulty that there is no consistent rule for all objective and subjective "sensors" to follow to assign their beliefs, I introduce an idea that I call the "Weighted Dempster-Shafer Evidence Combination Rule". This rule adjusts different sensors' voting effectiveness according to their specifications and historical performances. When the ground-truth is available afterwards or from an additional channel, this idea is further developed to what I call the "Adaptive Dynamically-Weighted Dempster-Shafer Evidence Combination Rule".
This innovative sensor fusion solution is proved to be effective with the prerecorded experimental data regarding users' focus of attention estimation in a small group meeting (done by Rainer Stiefelhagen), and many artificially generated experimental data.
You can find more detailed information regarding this work following this link. A copy of my dissertation is available here.
In the aging-aircraft skin inspection research in the Intelligent Sensor, Measurement and Control Lab in CIMDS, we paid special attention to the venerable coin-tap test method. In order to facilitate detecting signs of under-skin structural defects, we measured both the impact force and the resulting sound from coin-tapping, analyzed the 2-channel data to investigate their correlation-ship, examined various data fusion methods, and tried to present our results graphically to human inspectors. A brief project description can be found here.
The purpose of this research was to investigate the static and dynamic responds of tin oxide based solid-state gas sensors (most of which were supposed to be sensitive to oxygen-reduced gases) to fecal component gases and vapors. With the goal of building cheap "artificial noses" that could be used to monitor incontinence of elderly patients in nursing home environments, we studied the responses of seven easily available sensors try to find an unambiguous signature in their high-dimension behavior space. You can read our paper (pdf, 224KB) to get more information about this research.
In the CyberRAVEproject in the AML (Advanced Mechatronics Lab), the goal is to develop a swarm of battle-field small robots cooperating to fulfill a given reconnaissance or surveillance mission. For those tiny robots, sound information plays an important role in understanding their environment. My research focused on enabling the robots to recognize interesting sound signatures and trace the source object's movement. More detailed information regarding my research can be found here.
In the summer of 2001, I had a fulltime position in the Robert Bosch Research Center (RTC) North America, Pittsburgh, Pennsylvania. As part of the collaboration between the RTC and the CMU Robotics NavLab, my work was to help transfer the NavLab's sensor fusion software architecture package on Linux platform to RTC's Windows 2000 environment. My work also improved documentation and and investigated how to enhance the system with some context-aware computing features.
My summer internship in 2000 was in the Motorola Research Labs Applications Research Lab, Schaumburg, Illinois. Since Motorola Research Labs had a long-term research project to develop "killer" context-aware computing applications, my main task was to do literature research on sensing and sensor technology and to conduct a sensor technology roadmap study for the context-aware computing research project.
After finishing my M.S. degree, I worked in the Robot & Automation Research Institute, Beijing, China for over 8 years. I began as an Engineer Assistant in February 1987, was promoted to Engineer in February 1988, and to Senior Engineer in August 1993. I participated in the whole process of research design, development, and manufacturing support for the model HRGP-1A Painting Robot. I also participated in many other interesting projects there. You can find a brief description of my work here.
In the spring semester of 2000, I TA-ed the course 16-722 Sensing and Sensors, taught by Professor Mel Siegel. It is a Robotics Ph.D. Program perception core qualifier course, which covers lots of materials about the concurrent sensing and sensors' technology, from theoretical limitation analysis to practical implementation case study. Through reading, preparing answers and grading homeworks and exams etc. work, I got a deeper understanding of the main topics and wider exposure to this area.
In the spring semester of 1999, I TA-ed the 15-229 Multimedia Information Processing Course, taught by Professor Robert Thibadeau, Professor Roger Dannenberg and Professor Raj Reddy. This course taught students the basic principles of multimedia processing and trained them to use various image, audio, and video processing, and website-building software tools. As part of my work, I not only accumulated more experiences in processing multimedia information, but also improved skills and tactics to communicate with young undergraduate students.
If you ask people I work with, most likely they will tell you that I have a good temperation and am an easy-going guy. That is true. I do have an easy-going personality. I always try my best, but I am satisfied with what I get in life.
I grew up in a relatively harsh environment. I remember cold winters and shortage of food. But more than difficult times, I remember the kindness of my teachers and many other happy times. When I was very young I did not have many opportunities to learn and enjoy what I desired, such as playing music instruments in a marching band, but I take delight in the fact that my son has many more opportunities to realize his potential.
I believe that my childhood experience helped me to develop endurance, persistence, and perseverance. It also helped me to appreciate what I have now.
I spend most of my time working, but I also make time for physical excise. I have managed to work out 2 to 3 times a week for the last ~10 years. I play tennis in summer, and go jogging or lift weights in winter. My training goal was to run 10 kilometers in 50 minutes (approximately at a pace of 7.5 miles/hour). Now I can run 10 kilometers in 54 to 55 minutes (about 7 miles/hour) without much difficulty (heart rate <= 150 pulses/minute).
If I had time to enjoy myself, I would browse books and magazines, surf the Internet, listen to Chinese folk music or classic music, and play music instruments. I can play classic Chinese flute (made of bamboo with a resonant film).
|If you want to contact me, you are welcome to send me an E-mail at:|
Last modified: February 20, 2004