Biometrics takes a seat

By Kimberly Patch, Technology Research News

Your chair may be privy to a considerable amount of information, being so close to you so much of the time.

Researchers at Purdue University are taking advantage of the close relationship by adding pressure sensors to the seat and back of a chair so it can track how a person is sitting.

Currently, the chair can sense only static postures, but the researchers' eventual aim is a real-time tracking system that can sense how a person shifts throughout the day. This may eventually allow the chair to react in useful ways to how a person is sitting, or even to serve as a computer interface.

The chair's sensors are thin sheets of plastic embedded with circuits that give pressure readings from 4,000 separate spots, or pixels, on the chair back and seat. When someone sits in the chair, individual points of pressure at each pixel are recorded as numbers between zero and 255, resulting in a 3-D pressure map.

To enable the chair to determine how a person was sitting, the researchers had 30 people of different sizes sit in the chair in 10 distinct positions, including upright, leaning in one of several directions, slouching, or crossing either leg.

"The computer builds up a model of what [a given posture's] pressure should look like, and the population variation associated with it, because nobody sits down exactly the same way twice, and different people sit slightly differently," said Hong Z. Tan, assistant professor of electrical and computer engineering at Purdue University.

An image-processing algorithm used in face recognition did a principal component analysis to reduce the 4000 points of data to 15 potentially important features. Features are things like the position of the highest pressure point, the distance between sitting bones, or the total contact area, Tan said.

This data reduction process pinpointed both common features within a sitting position and the distribution of variation within those features. For example, "what's the mean pressure map for everybody sitting upright, and then what's the range of variation," said Tan.

Once the computer had the measurements for those samples, it was able to determine the posture of a person sitting in the chair by comparing pressure readings, she said.

"It's a very interesting first piece of work in this area," said Irfan Essa, assistant professor of computing at Georgia Tech. "I have not seen much work that other people have done on instrumenting chairs like this. In essence it's a first step in that direction," he said. In addition, the current work "could be used in a very limited sense [to] try to help improve... a person's posture," he said.

The chair's static posture abilities are a first step in a more ambitious dynamic tracking project, said Tan.

Once the chair can determine static postures, it can go on to show how a person moves in a chair. "[To] do the continuous tracking, we need to catch static postures very well and then model the transition. So this is the beginning portion of our work [toward] a continuous tracking system," said Tan.

To that end, the researchers are investigating two different methods. "One technique is the Hidden Markov Model, which people use to model speech. We think that could be used to automatically learn the pattern between different sitting postures when people switch between one and another," Tan said. The researchers are also looking into using neural networks in a similar way.

The researchers are aiming to make a user-independent system where "anybody could just sit down and the computer can start to track their posture," said Tan.

Potential practical uses of this ability range from ergonomics to driving safety to computer control.

For instance, it could be used as a tool to evaluate both the sitting habits of people and the ergonomics of chairs. A chair that keeps track of how a person moves all day could keep statistics and predict consequences, said Tan. "It can say something like, 'for 20 percent of today you're slouching. This is the typical pressure distribution when you slouch, and [this] part of your body is not properly supported when you slouch,'" she said.

As this type of technology matures there could be a lot of interesting possibilities like sensing if a person is falling asleep or getting tense while driving, said Essa. "This has a lot of potential," he added.

It could also serve as a sensitive human-machine interface, said Tan. "Here's an object that everyone sits in when they interact with computers... what if I use the chair [as a] control?"

For example, in an intelligent teleconferencing room application, "if I turn, maybe a remote camera pans left and right. If I lean forward maybe that means I'm interested -- it should zoom into whatever is the center of the scene. If I lean back maybe I want to get a shot of the whole room. This would a very intuitive kind of interface [that] you wouldn't even have to think too much about," said Tan.

And further into the future, a sensing chair may be able to at least assist in biometric identification. "It's pretty hard to fool a computer, the way the pressure distributes," Tan said.

For instance, from a chair's point of view, the parts of a person that produce the most pressure are the sit bones. And one of the very distinct features the chair can determine is the distance of the sitting bones, said Tan.

Even the current version of Tan's chair can tell with a fair amount of accuracy the difference between a male and female based on the size of the space between these bones. "If you are close to 16 cm [it's] very likely a female, if you're close to 14 cm it's very likely it's a male," she said.

A real-time posture tracking chair robust enough for practical applications is three to five years away, said Tan.

Tan's research colleague is Lynne A. Slivovsky of Purdue University. Slivovsky presented the research at the 2000 International Mechanical Engineering Congress and Exposition in Orlando, Florida on November 9, 2000. The research was funded by Purdue University and by the National Science Foundation (NSF).

Timeline:   3-5 years
Funding:   University, Government
TRN Categories:  Applied Technology; Human-Computer Interaction
Story Type:   News
Related Elements:   Technical paper, "A Real-time Static Posture Classification System," presented at the 2000 International Mechanical Engineering Congress and Exposition's Symposium on Haptic Interfaces for Virtual Environments and Teleoperation in Orlando, Florida, November 9.


November 15, 2000

Page One

Biometrics takes a seat

Oversize oddity could yield quantum computers

Switch narrows molecular-macroscopic gap

Tiny metal wires chart nanoelectronics

Chip techniques block power leakage


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.