THE HEART OF THE MATTER
a special program of the National Emergency Medicine Association (NEMA) 

Week: 564.6 Guest: Dr. Neil Gershenfeld, Head of Media Lab, MIT, Cambridge, MA Topic: Smart car seats help air bags Producer/Host: Steve Girard

NEMA: Airbags...from hero to front seat goat in the span of a couple of years. The auto safety devices have saved many lives in serious head on crashes...but as you know, concern of late is over tots in baby seats, and youngsters being put in danger when the airbag shoots open during a relatively minor accident. Deaths have occurred. Today we're with Dr. Neil Gershenfeld of the Massachusetts Institute of Technology...who's gotten involved with the issue in a roundabout way....tell us about it....

GERSHENFELD: Sure, it's a remarkable story. We didn't in any way start out looking to improve airbags. It's a really nice example of how the Media Lab works, and how industry and academia can work together. It actually starts way back with musical instruments. I run a physics group in the Media Lab that broadly looks at how physics and information are related to each other, and one of the fun application areas we've been working with is performance. We did a project with a composer at the Media Lab, Todd Machover, and YoYo Ma...developing sensors to generalize what a Stradivarius can do. And it was part of doing that project, we had to develop some new kinds of very sensitive devices, for example, to measure where his bow was...without interfering with it, but that could match his performance. Once we did that, we noticed some funny interactions between the fields we were sensing with and his body, and they didn't quite make sense. So, after those performances were done we came back to the lab and got interested in studying how fields in the human body interact. These are weak, low frequency electric fields. Now that's a very, very old idea. It goes back to the Theramin and early capacitance sensors, and is familiar in, say, elevator touch buttons. But what we noticed was, lurking behind what people thought was simple, was actually a very beautiful and very deep mathematical theory on how to not just measure a signal, but actually image...how to see with the fields. And that's what we've been developing. As that progressed, we realized we could do things like say make a table that could see in 3-D...so, it looks like an ordinary table, but it could find your hand....as a mouse, or a data glove, or something like that. And there's a lot of practical interest in that, because your desk is a cluttered work space, and rather than make room for the mouse, you always have your hand attached, so it would be nice to use that. Now, once we got that far, we then started having fun playing with these models of the body, and one of the things we did around that time was...we were talking to the magicians Penn & Teller. Now, mediums used to pretend to channel fields, to contact spirits that would make sounds. We realized we could do that literally, what we could do is make a spirit cabinet that would radiate a field out of the magician's body, and then we could sense that, and then we could have sounds and sights distributed in space and time around them. So that was just great fun, Penn & Teller used that for a series of performances, and then here's where the unexpected connection to air bags came. One of the sponsors at the Media Lab is NEC, and their President for Automotive Electronics, Phil Ritmuller, was visiting and he saw this Penn & Teller project....and he said, "Aha! Could you use that for the airbag problem"? And that was actually the first we had heard about it, but as you well know, the story was just about to break around that time of how if you fire an airbag with a rear facing infant seat, you can do great injury and even cause death for the infant. So the problem that Phil posed to us was, could we take our sensors and make a smart seat that could describe the occupant? So, the task of the problem is: the seat needs to be unobtrusive, robust and inexpensive, so it works in the automotive industry. And it has to distinguish a number of cases: it has to recognize if it's an adult sitting forward, in which case firing the airbag saves the adult's life...an infant in a seat facing forward, in which case firing the airbag saves the infant's life...an infant in a seat facing reverse, in which case firing the airbag can kill the baby....an out of position, or small adult, in which firing the airbag can cause injury...and finally, if there's just a bag of groceries on the seat, firing the airbag would waste a thousand dollars. The cost to replace it. And you really need a smart controller for the airbag.

NEMA: Now, the sensor you mentioned is something you're basically making...a seat with sensors in it rather than something that measures distances, for instance, between dashboard and occupant.

GERSHENFELD: That's exactly right...people had been doing a number of relatively obvious, but crude things. One example is a weight sensor, and you'd weigh the seat to make that decision. But the problem is the weight sensor can't recognize forward versus reverse, or it can't recognize groceries versus baby. Or people had tried to put optical sensors in the dashboard to look at the seat, but it's hard to tell by looking unless you put a super computer in the car, and even then it depends on how the car is illuminated. What was so interesting about this application of the devices we were developing was they go in the seat, and they help the seat to really image...to see the occupant. So the seat is really seeing in three dimensions... instead of sensing some crude property. And that's where the real power comes from. So what happened was at that point, even though we weren't an airbag research lab, it looked like such a nice application, we were very happy to veer off, and spend a few weeks building a prototype of a seat with the sensors. And then when we put in some test babies, much to our surprise, and pleasure, it really did work better than anything the industry had seen. Since then, we've been working very closely with NEC...which has now announced a product, and is working with the car companies to get it in. With this existence proof....that the airbag really could get smart...now the Highway Traffic Safety Administration and the industry is taking much more seriously the possibility that in the fairly near future, you can really have smart airbags, that are intelligent and respond to the occupant.

NEMA: It's amazing to see something like this go from idea to seminal stage to production, installation, use....all within a couple of years...

GERSHENFELD: That's right...now, what's accelerating it is the enormous demand in the industry. They know they need the solution. But for me, actually...this is gratifying. And I should mention right around the time this whole story was happening, I had twins... one of whom I'm giving a bottle to right now...and it was very gratifying to then realize that in this satisfying confluence, I was doing something that would help them and their safety in my car. But even as satisfying as that is, what I really like about the project is what it points to beyond that. This is really just the first step in making the car more responsive to the occupant. I co-direct a consortium at MIT called Things That Think...what does it mean for intelligence to go out in the world around us. And one of the projects there, run by one of my colleagues, Professor Ross Picard, is looking at what we call affective computing. And the idea is...you're affective state is...are you tense? or confused? Or drunk or happy or sad? And that's real, that's falsifiable...that's not pseudo-scientific. And how people deal with each other depends crucially on that. You know, if I was mad, you'd deal with me very differently than if I wasn't. But machines don't have access to people's affective state, and so in a sense they're emotionally handicapped. Now, once a seat like this is in the car, with the kind of mathematics that Roz' group is developing, you can then start to analyze how a person moves in the seat, and start to look at patterns of things like drowsiness or drunken behavior....so that then the car could start to respond appropriately. Let's say if you're falling asleep, it could turn on cold air and help wake you up.

NEMA: Does this technology have any implications for the medical community?

GERSHENFELD: Things That Think is turning out to have a much stronger focus on medical problems than I expected. One of the most important sponsors is Becton Dickinson, for example, and they made a billion syringes last year, so it is very important for the syringe to become smart. Let's say diabetics are notoriously bad at medical monitoring...and so if you really want intelligence at the lowest levels of the medical system, in even the consumables... like the syringes...they have to cost pennies. You can't afford to put chips everywhere. We're doing a lot of work on remote sensing of smart materials... on using interesting materials to replace chips. So that then you can have things like a smart syringe that can measure who got how much of what? But all you've done to the syringe is build interesting materials into it. And so, one of the big pushes in Things That Think is by using techniques like that, push medical monitoring much much closer to people, so that rather than something that's done infrequently with expensive devices, it's something that can be done continuously and routinely with the objects around you.

NEMA: Explain the workings of that application to me....

GERSHENFELD: Well, a simple example are shop lifting tags...another one of our sponsors is Sensormatic. And they make, for example, when you buy a compact disc, there's a little, white, squishy thing on the back that detects if you're stealing it. That thing is actually a magneto elastic metallic glass... that's what it's called. And when you excite it with a weak, or a low frequency magnetic field, it essentially sings...it resonates in a very characteristic pattern that let's you detect it...so that's how it can determine that you're stealing something. Well, what we found is that the physical mechanism that lies behind that sort of process... for example, these magnetic glass resonance's...and there are a number of other objects that have a similar property. You can encode information in the resonance structure, so by placing one of these materials in proximity to a syringe, it's response depends on the liquid in the syringe. So that then, a reader elsewhere can detect, 'this is in the vicinity'...and not just simply detect something as simple as theft, like happens at the check out counter, but could actually measure the liquid volume remotely, with the syringe. And I should mention we're doing a lot of work also in developing consumer versions of nuclear magnetic resonance, to do non-invasive chemical monitoring.

NEMA: All that with a babe in his arms...our thanks to Dr. Neil Gershenfeld of MIT. And as an aside...his twins have a high tech birthday... October 10....if you're not digitally challenged, that's 1-0-1-0.

SPOT: 15 years in the prevention of heart disease, stroke and trauma - The National Emergency Medicine Association. This show is just part of what NEMA does. We send out millions of pieces of prevention information to people around the country, give grants to organizations in research, public information and emergency services, and have been instrumental in the creation and expansion of the Chest Pain Emergency Room movement. To play a role, call 800-332-6362.

NEMA: For more information or transcripts of our programs, check into our home page:

www.NEMAhealth.org (that's n-e-m-a)

...or use your search engine to find The National Emergency Medicine Assn. I'm Steve Girard at The Heart of the Matter.