Human Design Research Group

Sumit Basu, Ph.D. Candidate, EECS

Sumit Basu

My research is in machine perception, machine learning, and human-computer interfaces. I am currently working on "Conversational Scene Analysis:" using low-level audio and visual cues to better understand human-human and human-computer interactions. It seems that we can understand something about the nature of a conversation without even knowing the language, and my goal is to attain this sort of understanding automatically. I am developing statistical models to determine, classify, and predict the "state" of a conversation. I am also investigating how manipulating the channels of interaction can affect the conversational state. PhD expected June 2002.


Conversational Scene Analysis	The Facilitator Room	Smart Headphones	Speech/Pitch Detection	Deformable Models	3D Head Tracking

Tanzeem Choudhury, Ph.D. Candidate, MAS

Tanzeem Choudhury

I am interested in systems that can learn how humans behave and interact with their environment and with each other. These systems will then be able to use this knowledge to interact with and take actions on humans to achieve some desired effect. In my research I am exploring the possibilities of using three sets of tools -- (i) perceptual sensors that give us features of human behavior and interactions, (ii) computational models that can extract patterns/structure from these features and learn behavioral models, and (iii) interdisciplinary approach and insight into modeling and understanding complex interactive networks -- to understand the structural relationships among interacting groups of people in smart environments, and ways to take actions which bring about some desired effects or achieve desired goals.

Details coming soon ..


Learning Social Networks	The Facilitator Room	Speaker Detection using the Smart kiosk	Facial Expression Understanding	Multi-modal Person Identification

Brian Clarkson, Ph.D. Candidate, MAS

Brian Clarkson

In the 70's there was a relatively obscure conceptual artist named On Kawara. He was obsessed with time and the mundane events that mark its passage. His works explored the kind of day-to-day events that tend to fall between the cracks of our memories. For years, everyday Mr. On would record the exact time he awoke on a postcard and send it to a friend or create lists of the people he met each day or trace on maps where he went each day. His work raises a few interesting questions. If we had consistent records of some aspects of our day-to-day lives over a span of a lifetime, what trends could we find? What kinds of patterns or cycles would reveal themselves? Interestingly, we wouldn't need highly detailed memories to find these trends and patterns, just a consistent sampling in time. My dream is to build a device that can capture these life patterns automatically and render them in a diary-like structure. I am certain this device could help us understand our present state in the context of our past experiences.


The "I Sensed" Series: Modeling an Individual's Day-to-Day Activity	The Facilitator Room: Modeling Human Interactions

Nathan Eagle, MS Candidate, MAS

Nathan Eagle

Machine learning, ubiquitous computing and appropriate technology are among my primary academic interests. Currently I'm collaborating with other students in our group on a project involving the design of a classification algorithm that can deduce meaning from data sets of computer-transcribed conversations. My ubiquitous computing experience includes creating applications for networked, location-aware handheld computers based on the infrastructure of Project Oxygen at the Laboratory for Computer Science. Within the appropriate technology domain, I am particularly interested in developmental entrepreneurship and the application of innovative machine learning techniques to rural medical diagnosis. My current research involves leading a portion of Ca:sh, a village-health project within Media Lab Asia.


Conversation Context Extraction	Project Oxygen	Ca:sh - Community-based, accessible & sustainable health	Fulbright Research in Nepal	Drowsiness Detector (Stanford MS EE Project)

Tony Jebara, Ph.D. Candidate, MAS

Tony Jebara

I am investigating new ways of teaching computers and agents what to do. Instead of programming them we let them watch a phenomenon such as two humans interacting and learn behaviour. To this end, I have developed perceptual tools such as MeshTrack which detects and tracks faces in 3D as well as recovers 3D deformable models in real-time. In addition, I have developed a framework called Action-Reaction Learning which implements statistical prediction techniques to learn from such perceptual measurements. By watching people interacting it can understand and acquire the behaviour exhibited. I am also interested in human augmentation and wearable computers. Some projects include wearable face recognition and wearable billiards enhancement.


MeshTrack 1	MeshTrack 2	MeshTrack 3	Action-Reaction (clap)	Action-Reaction (scare)	Action-Reaction (wave)	DyPers

Vikram Sheel Kumar, MD Candidate, HST

Vikram Sheel Kumar

My research focuses on building a class of preventative medicine using portable and wireless tools that empower patients with insight into their conditions and appropriate community support. While our current health systems are suitable for tackling acute care, patients carry the onerous burden of managing chronic conditions where healthy behavior is of essence, and regular monitoring is necessary. By giving feedback on their continually changing physiology, I am studying how patients can learn to develop accurate mental models of their conditions. Two such projects are DiaBetNet, a community-based wireless game for Type I diabetic children, and HiRoller, a self-monitoring aid for patients with bipolar disorder. In a project through Media Lab Asia called Ca:sh, I am are studying how handhelds can be used to provide meaningful representations of data to assist rural health workers in India in promoting healthy behaviors amongst mothers and children.


DiaBetNet	HiRoller	Ca:sh - Community-based, accessible & sustainable health

Steven Schwartz, Research Scientist

Steven Schwartz

As a Research Scientist my role is to provide siginificant contributions to researchers in our group. Pushing the edge of the envelope for Human Centered Information Systems through development of advanced wearable computing platforms is what I do. Since 1999 my research projects have led to the design of computing architectures and infrastructires that support computing environments compatible with clothing and intended for a wide variety of applications. The type of clothing used for these wearable computing experiments has ranged from simple mesh vests to spacesuits used by US astronauts during spacewalks. The challenge ahead is to develop methods for developing these information devices into productive and pleasing tools that do not burden those who use them.


Smart Vest	MIThril	WearARM	WearSAT