Codes software that makes handheld computers handier.
Lorraine Wheeler is forever moving on to the next big thing. As a 22-year-old software engineer at GE Medical Systems, she devised a process to improve the signal-to-noise ratio of magnetic-resonance imaging. Her patented innovation helped clarify results for doctors who rely on such images to diagnose disease. Then, with a mere $5,000, the 27-year-old Wheeler founded Actual software above a pizza shop in Andover, MA, and created MultiMail, one of the first e-mail programs for Palm handhelds. With MultiMail’s, business travelers could use their personal digital assistants to send and receive e-mail on the road. Upon its release in 1998, MultiMail’s free version was downloaded 1,000 times per hour, and the full-featured version was so successful that Palm eventually bought Actual Software for $4 million. In 2002, Wheeler founded Botzam. The startup , housed in more comfortable commercial digs in North Billerica, MA, is developing new applications, including software that lets users back up their handheld data on removable storage cards. “I’m interested in pure innovation,” Wheeler says. “I follow the market wherever it moves.”
Gives unmanned reconnaissance planes insect vision.
If you followed the recent wars in Afganistan and Iraq, you’ve likely heard of unmanned aerial vehicles such as the U.S. Army’s Predator. These craft allow hostile-area reconnaissance with no risk to pilots, who use Global Positioning System-based navigation to guide them remotely. Precision guidance, though, can be difficult from afar. Enter Geoffrey Barrows, founder and president of Centeye, a three-year-old Washington, DC, company that develops “bio-inspired” microelectronics. Centeye is commercializing optic-flow sensors, chips designed to help unmanned aerial vehicles navigate autonomously by endowing them with the kind of depth perception exhibited by flying insects. The chips, which Barrows developed between 1997 and 2000 while working for the U.S. Naval Research Laboratory, compare objects’ rate of movement through the visual field to deduce their relative distance. With contracts from the U.S. Defense Advanced Research Projects Agency and the Naval Research Lab, Barrows is working to reduce the sensors’ weight to just a few grams; he aims to deploy them in small, fast-moving robot planes within three years.
Devises powerful tools for assembling and analyzing genomes.
Science has fascinated Serafim Batzoglou since he was 12, when he watched Carl Sagan on TV, at home in Athens, Greece. Today, Batzoglou is unlocking the mysteries of the genetic structure of organisms, and he has recently led development of two important programs for deciphering genomes. Arachne, released in 2002, was the first widely available tool for assembling full genomes from fragments of genes. Researchers used it to assemble the mouse genome, creating a blueprint for studying the human genome (mice and humans share 95 percent of their genes). Scientists also used Arachne to sequence another scientific workhorse, the fruit fly. The second program, Rosetta, helps compare the human and mouse genomes. Although Batzoglou’s algorithms draw great praise, he says he’s simply lucky his programming skills can be applied to Biology. Now a Stanford University assistant professor of computer science, Batzoglou heads development of Lagan, a software tool for aligning long sequences of DNA so scientists can compare the genetic structures of humans and other creatures. Batzoglou is not a biologist, but his work may help rewrite the book of evolution.
Constructs robots whose expressive faces convey humanlike emotions.
People often ask Cynthia Breazeal, director of the MIT Media Laboratory’s robotic-life group, whether robots will take over the world. “I’m like, ‘Oh, go see a movie,’” she laughs. Nonetheless, there’s something Hollywood about Breazeal’s work. She builds expressive robots that exhibit socially appropriate emotional responses, attracting media attention as well as advancing artificial-intelligence research. For her doctoral thesis, Breazeal constructed Kismet, a bright-eyed mechanical head that reacts to human voices, movements, and expressions with smiles, frowns, and raised eyebrows. Her latest robot, Leonardo, a collaboration with the Stan Winston Studio, of movie special-effects fame, is a 75-centimeter-tall creature. The furry bot’s 60 small motors produce fluid movements and subtle facial expressions; it also has pressure-sensitive “skin”, microphones, a speech synthesizer, and camera “eyes” that track people’s faces and gestures. Unlike other robots, whose actions are driven by programmed routines, Leonardo learns tasks by assessing humans’ expressions and imitating their movements. Breazeal calls it “the most expressive robot today,” and because she believes “socially intelligent” robots could become actors, or helpers for the elderly, she is conducting studies of human-robot interaction. Her lab is also helping NASA build a “robonaut” space assistant that might one day perform maintenance tasks in space.
Pioneered software that delivers Web files quickly, anonymously.
Ian Clarke didn’t set out to ignite a debate about Internet free speech and copyrights. But that’s what happened in 2000 when he released Freenet, a free program that shuffles files among Internet-connected computers, enabling people to store and retrieve data easily and anonymously. Maybe is was the fact that , unlike Napster and other peer-to-peer file-sharing systems that rely on observable central indexes, Freenet makes it almost impossible for sensors or copyright owners to trace files. Maybe it was the uncompromising philosophy the County Meath, Ireland, native readily shared: “You cannot have freedom of communication and enforce copyright law.” Yet Clarke says he designed Freenet mainly to test a technical idea he hatched at the University of Edinburgh undergraduate: that a network capable of replicating documents and storing them in multiple locations would bring files closer to frequent users, speeding their delivery. Enthusiasts have since downloaded close to two million copies of Freenet. In 2001 Clarke raised $4 million to start Uprizer in Santa Monica, CA, which sells corporate information management software based on Freenet. He’s now cofounder and CEO of Cematics, also in Santa Monica, a peer-to-peer product development and consultancy company.
Designs architectures needed to build practical molecular computers.
Caltech computer science professor Andre DeHon is changing the architecture of integrated circuits- and it needs changing. Conventional circuit elements are nearly as small as they can be, so materials scientists are fabricating nanowires 10 times smaller to replace them. DeHon described how to chemically modify the ends of nanowires so that the much larger wires used in conventional circuitry could address them individually. He then showed how to arrange such wires into working circuits, even within the limitations of existing lab techniques, and he has developed an architecture for building a general-purpose computer from them. Moreover, his processor can be reprogrammed to perform different computing tasks that would ordinarily require distinct architectures. This extends the work DeHon did as an MIT graduate student on reprogrammable semiconductor chips, a technology commercialized by several undergraduates he supervised, whose company, Silicon spice, was acquired by Broadcom for $1.2 billion. “Andre’s work is striking,” says Harvard University nanotech pioneer Charles Lieber. “After reading it one must say, ‘Yes, this idea of molecular computing is indeed a real possibility, not just hype.’”
Works to improve quantum computers so they can speed drug design and perform other massive computing tasks.
Daniel Gottesman is increasing the chances of building quantum computers that could solve certain large-scale computing problems much faster than current-day machines. Atomic particles can exist in many states at once, and if those states can be controlled, they could provide far more computational power than the simple on-off states of electronic bits. The problem is that quantum computer elements are highly susceptible to errors. While a grad student at Caltech, Gottesman helped develop a systematic method for correcting those errors, thereby stabilizing quantum computers. His methods are used worldwide by researchers designing the first rudimentary quantum machines. Now a research scientist at the Perimeter Institute, a nonprofit research facility in Waterloo, Ontario, Gottesman is writing protocols that will use quantum mechanics to provide extremely secure data encryption codes- in part because quantum computers will be powerful enough to break current encryption codes. Although practical quantum computers are still years away, Gottesman says their computational speed could “revolutionize” the design of drugs and new materials such as high-temperature superconductors.
Fabricates three-dimensional integrated circuits that could vastly increase computer power.
“Remarkable” is how Stanford University professor Cal Quate describes his onetime PhD student Kathryn Guarini. In the mid- 1990’s, Guarini and Quate developed a lithography technique that let engineers pattern integrated-circuit features smaller than 100 nonometers, which packed more circuits, and thus more power, onto chips. Since 1999, when she joined IBM Research, Guarini has led Big Blue’s development of three-dimensional circuits- a semiconductor frontier. For decades, designers have made transistors smaller but have continued to place them side by side in a single layer. Guarini’s techniques can stack transistors in two or three layers, vastly increasing the number that fit on a chip. She has also shortened the metallic connections between transistors, accelerating processing speeds. Challenges remain, including how to limit the heat a 3-D chip produces. But colleagues predict Guarini will prevail. Says fellow IBM researcher Philip Wong, “Whatever she touches, she turns to gold.”
Sparked Microsofts change to .Net.
In 1991, Vic Gundotra, a 21-year-old programmer at Microsoft’s Washington, DC, office, showed some of his demos to Steve Ballmer, then Microsoft’s head of sales (and now CEO). Soon after, Gundotra was moved to Microsoft headquarters in Redmond, WA. There, Gundotra and colleagues ran the skunk works project that resulted in a new architecture for supporting company software applications across many kinds of computers; now knows as .Net, it won out over several established projects. In 1999, Gundotra also wrote the code for the network software that underlies .Net- which crystallized “Web services” for Ballmer and Bill gates/ The code sparked Microsoft’s next PC operating system, Longhorn (due in 2005) will be driven by ,Net. Gundotra is a born proselytizer for technology: he uses a tablet PC to take notes during church. He doesn’t know if he’ll stay at Microsoft forever, but he says, “If you want to change the world, it’s much easier to do it at Microsoft.”
Invented inexpensive rocket-based surveillance systems.
Andrew Heafitz acquired his first patent- for a high-speed shutter on a balsa wood camera- at age 19. Since then he has designed electric vehicles, invented a low-cost rock engine, and developed a plant growth experiment that flew on two space shuttle missions. His MIT undergraduate thesis centered on a remote-control aerial-photography system he built using a Nikon camera and a helium balloon. While he was a mechanical engineering master’s student at MIT, he installed a tiny off-the-shelf video camera inside the nose of a small rocket he’d designed. He called the device TacShot and in 2001 launched a company of the same name in Cambridge, MA. The 350-gram self-propelled rocket travels to heights of 100 meters, radioing images of the earth to a ground-based laptop. Heafitz’s software assembles the pictures into a 360-degree mosaic of the launch site. TacShot could be carried in a soldier’s backpack and deployed for aerial reconnaissance or used to assess damage at disaster sites. Heafitz won the 2002 Lemelson-MIT Student Prize and invested part of the $30,000 award in his company. With a U.S. Air Force contract to conduct proof-of-concept demonstrations, Heafitz has TacShot- and his career- on a successful trajectory.
Devised software that roots out security threats to a networks operating system.
When investors backed Sana Security in San Mateo, CA, they took out $1 million in insurance on chief scientist Steven Hofmeyr, because his ideas about computer network security were their most valuable asset. Sana Security (“sana” is Esperanto for healthy) treats intranets and local-area networks much the way the human immune system treats the body: when abnormal activity occurs, Sana software recognizes the problem, isolates and analyzes it, then attempts to fight it off. Network security typically relies on human administrators who often arrive too late to stop intrusions (in January the Slammer virus reportedly infected 75,000 machines within 10 minutes). Other tools, such as software updates, arrive after the fact and can be expensive. Sana’s Primary Response product resides on each computer within a network and continuously interrogates vulnerable applications, such as e-mail programs, at the operating-system level. When it detects abnormal behavior- hackers looking for loopholes, say- it stops the activity, then notifies a system administrator. Within two months of the program’s release, several large organizations became customers, including NASA.
Engineers tiny sensors that can be spread like crumbs around a battlefield or factory.
Mike Horton is yanking the wires out of your life. His company, Crossbow Technology in San Jose, CA, makes truckloads of solid-state wireless sensors no larger than pagers. Within a year, he promises, they will be as small as bottle caps. In some cars, Crossbow’s sensors already report acceleration and engine pressure data to engine controllers. In aircraft, they are replacing the larger, less reliable, airframe-mounted mechanical gyros that correct for altitude, roll, and drift. But Horton, who holds two patents, plans to storm the wireless field with “smart dust”: sensors the size of rice grains. Sprinkled around a battleground or office building, the sensors would network themselves and analyze the environment, checking for vibrations that could indicate the passing of an enemy convoy, or airborne chemicals from an industrial spill. Having raised $13 million from Intel and other investors, Horton has partnered with mentors at the University of California, Berkeley, to write software for the motes, even as he works on reducing their size and cost. Horton believes billions of unnoticed dust sensors cold transform society: “It’s a quiet revolution,” he says.
Writes programs that more intelligently guide actions of robots.
Electrical engineer Ayanna Howard sees a future in which humans and machines work together to explore new terrain. Today’s planetary rovers- suitcase-size robots that move on wheels- must be remotely controlled by human operators, a labor-intensive and imprecise process. So Howard, who weeks at NASA’s Jet Propulsion Laboratory in Pasadena, CA, developed artificial-intelligence software that mimics the decisions humans make about where to land spacecraft and how to navigate robots safely. Loaded onto a rover’s computer, Howard’s software can process information from maps and video cameras and automatically find a safe path between two points, in real time. The rover can thus avoid craters instead of trying to negotiate their edges- at great risk to multimillion-dollar equipment. Rovers equipped with Howard’s software are now used by forest rangers in Idaho to map loggings shattered by earthquakes or bombs. In 2001, Howard won the Lew Allen Award, JPL’s highest honor for leadership and innovation.
Integrates photonics and electronics on chips to speed telecommunications.
As a 16-year-old South Korean émigré living in a gang-ridden Los Angeles area, Kevin Lee had to face the threat of daily violence. That situation, he says, “gave me drive.” In 2001, that drive prompted him and fellow TR100 honoree Desmond Lim to start LNL Technologies in Cambridge, MA, to revolutionize telecommunications. Frustrated by slowdowns that occur when fiber-optic signals are converted to the electrical signals that go into homes and offices, researchers had long tried to fashion circuitry that would manipulate light and speed up telephone, data, and video services. During his PhD studies at MIT, Lee crafted tiny optical routers and connectors that fit on a single chip. Building on that work, LNL has made prototype chips that integrate up to 10,000 photonic and optoelectronic functions within one square centimeter. The optical chip can be mass-produced and , the founders maintain, is far more powerful than other companies’ prototypes. “Optics is the industry of the future,” says Lee, vice president of engineering. “I want to be part of making that happen.”
Develops high-volume manufacturing lines for making optical chips into commodities.
Singapore-born Desmond Lim likes to convert scientific advances into marketable products. As cofounder of Cambridge, MA-based LNL Technologies with fellow TR100 member Kevin Lee, Lim aims to transform the company’s groundbreaking optical-chip prototypes into inexpensive commodities that will upgrade data transmission over phone lines and the Internet. To that end, chief technology officer Lim has designed a fast-turnaround process for prototyping optical components such as waveguides, filters, and multiplexers, and he has developed practical, high-volume manufacturing lines to make chips for the telecommunications industry. When LNL demonstrated its flagship chip in December 2002, the Wall Street Journal said it “could one day become the Pentium processor of light signals.” The company is now setting up field tests and talking to manufacturers about building LNL’s chips. “We are leveraging what is known about standard silicon chips and putting our ideas on top,” Lim says. “We are past the inspiration phase and are now in the sweat phase: moving toward mass production.”
Designed an automated tractor steering system that is saving farmers bushels of money.
As Michael O’Connor was completing his Stanford University PhD in aeronautics and astronautics, he realized that the Global Positioning System so central to his studies could save farmers millions. O’Connor designed a system that could guide a tractor, combine, or harvester to within three centimeters of a specified path. Upon graduating, he cofounded IntegriNautics in Menlo Park, CA, to commercialize it. The system consists of a small box installed in a tractor cab to control its steering mechanism, plus a touch screen, antenna, and GPS instrumentation. Once the farmer programs the path on the screen, the setup allows for hands-free operation that is more accurate than manual steering- and more economical, since even slight deviations can cost thousands of dollars in wasted fuel, fertilizer, and pesticides over many hectares. Although the farmer has to sit in the cab to watch for unpredictable obstacles, like animals, he can do other work while there. The system also allows operation during fog or at night. O’Connor has sold more than 400 of the $45,000 AutoFarm units. “Half the new tractors in the U.S. will have this technology within ten years,” says Stanford professor Brad Parkinson, a GPS pioneer who supervised O’Connor’s thesis.
Delivers "spotlights" of sound for use in concerts, museums, and automobiles.
Joe Pompei founded Watertown, MA-based Holosonics on what many acoustics experts called a crackpot idea: directing sound in a narrow beam. But Pompei says his attitude was, “this is too cool not to work.” He developed the initial technology in 1999 at MIT, the only school that gave him the green light on his PhD proposal. Within three months, he had a prototype of Audio Spotlight. The system includes a processor and a pizza-pan-sized transducer. First, the processor compresses the ling wavelength of an audible frequency into millimeter-long wavelengths of inaudible ultrasound. Then the transducer transmits the now tightly focused signal. In the air, the ultrasonic waves begin to elongate and regress to the original audible frequency, but only within the confines of a narrow beam. At distances of up to 200 meters, listeners in the beam’s path hear sound clearly, but those mere meters away from them do not. Dozens of groups have adopted the system: rock band U2 is experimenting with it to direct music at concerts; it’s used in museum and trade shows; and DaimlerChrysler installed it in Maxxcab truck models to provide personalized audio zones for passengers.
Makes simpler, more powerful animation tools for novices and professionals.
As an undergraduate, Jovan Popovic found that animated illustrations of math-problem solutions helped him master complex concepts. But that animation is tedious and expensive. So Popovic, who grew up in Yugoslavia and is now an MIT assistant professor, set out to make it easy and affordable for neophytes and professionals alike. He developed software that works like the drag-and-drop feature in word-processing programs. To show, say, a hat flipping through the air and landing on a coat rack, the creator simply mores the hat with a computer mouse, and Popovic’s program generates the appropriate animated movement. “He has made significant progress toward solving one of the key problems in animation: to make it look natural and control it at the same time,” says Andrew Witkin of Pixar Animation Studios. The entertainment industry should benefit, but Popovic wants to reach teachers, children, and educational filmmakers. He isn’t sure whether to license his software or start a company to develop it. Up next: programs that create animation in response to verbal commands.
Tailors Internet application to cell phones.
Microsoft’s original Internet Explorer development team consisted of one person: Thomas Reardon- or just “Reardon,” as he’s generally known. As Internet mania grew, so did Reardon, from programmer to program manager, sifting through hordes of unproven technologies and emerging standards, deciding which to adopt or reject. His work culminated in Explorer 3.0, the first Microsoft browser sophisticated enough to compete with Netscape Navigator. Reardon spent the next four years working with standards bodies, driving Microsoft’s move away from proprietary technologies and toward the open standards that enable software interoperability on the Internet. Today Reardon is a general manager at cell-phone software supplier Openwave in Redwood City, CA, where he’s waded into the middle of the next browser war. “We’re trying to kill this mentality that smart phones are just PCs ported to cell phones,” he says. Instead, he is directing Openwave toward software tailored to just the applications customers seem to want- such as picture messaging and the short-message service. That strategy has paid off: more than 80 percent of U.S. call phones now use Openwave’s wireless-Web browser, and the company expects sales this year to top 180 million units.
Employs simulations of human movement to create realistically animated characters.
It might be surprising to find a biologist pushing the frontiers of computer animation. But Torsten Reil is bringing cheaper, lifelike digital characters to video games and films. As a doctoral researcher in neural systems at the University of Oxfrd, Reil programmed computer simulations that mimicked human and animal movement, and in 2001 he cofounded NaturalMotion in Oxford, England, to commercialize that work. To create characters that move realistically, conventional animators draw extensive series of frames that are played back- repetitively- in set sequences. But Reil wrote software that an animator uses to program a nervous system for a character he or she draws just once. The code makes the character’s body obey the laws of physics and react automatically to changing on-screen situations. NaturalMotion’s first product is already saving game developers and visual-effects companies thousands of dollars by accelerating animation, Reil says. Look for his characters in the upcoming Hollywood epic Troy. Reil recently won a grant from the British government to model the gaits of children with cerebral palsy, to help doctors determine the neurological basis of the disorder.
Built large, bright, organic video displays using materials dismissed by contemporaries.
Heike Riel left a furniture-making apprenticeship to study physics. A PhD later, she has built the world’s largest full-color display that uses organic light-emitting diodes (OLEDs)- paving the way fro a new generation of vivid-color, affordable , flat-panel televisions and computer monitors. Her 20-inch screen is brighter and more energy efficient than any other screen on the market. Cell-phone displays and other small screens have used OLEDs for streaming video, but larger-scale applications have proven elusive. In a large screen, each of millions of light-emitting pixels requires several transistors, and the transistor matrix had been difficult to manufacture uniformly. Researchers had steadfastly tried to improve the polycrystalline-silicon transistors, claiming that the alternative- amorphous silicon- would break down at the high currents needed for pixel emission, But Riel and her colleagues fashioned the OLEDs at IBM’s Zurich lab so that inexpensive amorphous-silicon transistors drew less current and, therefore, remained particular, tinkered with the width of the pixel’s thin organic layers, allowing significantly more light to be emitted. “People didn’t believe it could be done,” she says.
Programs computers to recognize objects the way the human brain does.
The human brain can recognize a dog as a dog, but teaching a computer that trick is daunting. As an MIT postdoc, Max Riesenhuber researched the brain’s object recognition processed, then led a team that wrote software to mimic them. Called HMAX, the program is accurate enough to save physiologists time and money in studying brain disorders. To test the model, researchers might show it an illustration of a composite creature, such as a catlike dog. HMAX categorizes the animal’s features as more or less catlike or doglike, sums those probabilities, and issues judgments remarkably consistant with human subjects’. Now, scientists are using HMAX to craft better experiments to help explain brain disorders like prosopagnosia- the inability to recognize faces. Unraveling such afflictions is Riesenhuber’s main goal. But his software also advances computers’ ability to recognize objects , a key to artificial intelligence. HMAX might even help recognize satellite images. Riesenhuber, founder of MIT’s Motorcycle Club, is also a principal of GeoPhoenix in Cambridge, MA, which markets a handheld computer interface that can access content by zooming and panning, helping users navigate small screens.
Helps entrepreneurs in emerging nations turn innovations into business.
Lindo Rottenberg believe that startups are the best way to create jobs and stimulate growth in economically struggling nations. To prove that contention, she cofounded Endeavor, a nonprofit that helps entrepreneurs in emerging countries access networks, training, and everything else they need to convert innovative ideas into companies. Endeavor invests no money itself. Instead, CEO Rottenberg and five other New York-based staffers coordinate 24 employees in Argentina, Brazil, Chile, Uraguay, and Mexico- and soon in South Africa. Those employees encourage prominent local businesspeople to find funding, give advice, and otherwise open doors for entrepreneurs. In the past year Rottenberg’s group has developed Web chat rooms where executives and entrepreneurs can compare notes. Endeavor claims that since 1997 is has supported 121 companies, which have generated more than 9,000 jobs and $363 million in revenue. Startups that are profitable within a few years often reward Endeavor with a small percentage of revenues or equity. Endeavor’s notable technology startups include Patagon.com, an Argentine online finance pioneer that was sold for $700 million, and Taho, a Brazilian wireless-communications company. “People worldwide with good ideas are calling us every day,” Rottenberg says.
Fashions photonic circuits that could speed voice and data to homes.
Telecom networks are half-hare, half tortoise. Conversations and data blaze down fiber-optic cables but slow to a crawl when they encounter electronic switches at network junctions. Photonic crystals, which can manipulate photons much as semiconductors manipulate electrons, are the best hope for clearing such roadblocks. While others are exploring them, Ted Sargent is close to building practical devices. Sargent came to the University of Toronto as a grad student in 1995, joined its faculty in 1998, and in three years was awarded a coveted Canada Research Chair position. His chief advance is a process to specify and guide the growth of photonic crystals- a mix of electrochemistry, microchip fabrication, and holographic printing. Sargent etches a holographic pattern into an electrically conductive film that coats a glass plate. Then he electrifies the film, generating a holographic “tractor beam” that attracts latex beads and stacks them into a photonic crystal. Sargent has fashioned hair-thin rows of crystals that could act as circuits, and Nortel eagerly awaits proof they can manipulate photons. If so, photonic crystals could speed up the telecom network, including the “last mile” of cable to homes.
Serves up customized audio and video gems.
A year after leaving Swarthmore College, where he helped fellow undergraduates engage in scientific communication by cofounding the Journal of Young Investigators, Tim Sibley has an insight about a related form of communication: conferences. “What scientists are truly interested in,” he explains, “cold be just 20 minutes of one lecture out of a hundred hours at a conference.” A simple way to find relevant morsels within audio or video conference recordings would be a boon. So the mathematics and physics major secured $2 million form the National Institute of Standards and Technology to start StreamSage in Washington, DC. Sibley uses computational-linguistics techniques to automatically create a searchable index of an audio or video recording and find material relevant to any given topic. Today NASA and Harvard Medical School use his programs to provide access to streaming-video archives. Other customers might include media companies, which seek better ways to manage digital video. Next, Sibley plans to enlist speech-processing technology to stitch together personalized audio newscasts from the Web’s welter of news. “More than just an audio Google,” Sibley says, this system will put items in context.
Transforms computers ability to recognize human faces.
The trouble with facial-recognition software is faces. A computer probably wouldn’t recognize Osama Bin Laden if the lighting or his expression didn’t match those in a database. Using a specialized form of mathematics called tensor algebra, Alex Vasilescu- a research scientist at New York University and a computer science PhD candidate at the University of Toronto- has developed breakthrough recognition algorithms called TensorFaces. The algorithms promise to enhance computers’ ability to match multiple characteristics of a face in way that overcome vagaries of shading , angle, or expression. The U.S. Department of Defense is funding Vasilescu’s research because of its antiterrorism potential. She can also generate “motion signatures” on the basis of a person’s gait- walking, dancing, or running. Such gait changes can indicate the onset of certain illnesses. Finally, Vasilescu’s algorithms could help computers impart realistic textures to images of hair and fabric, which could yield better graphics for video games and movies.
Builds brain-imaging machines that are faster and cheaper than magnetic-resonance imaging equipment.
Tsuyoshi Yamamoto is steering his way through the brain. Since joining Hitachi’s Advanced Research Laboratory in Hatoyama, Japan, in 1997, he has codeveloped an optical-topography system that measures changes of concentrations of hemoglobin in the brain, providing insight into how neurons process language, images, and movement. Yamamoto’s device, which is easier to use and cheaper than magnetic-resonance imaging (MRI), beams light through the skull and into the cortex, the brain’s outer layer. By charting the intensity of the light reflected back by the cortex, the machine gauges changes in blood cell concentrations- indications of brain activity. The instrument, which looks like a beauty salon dryer, is less motion sensitive than MRI equipment, so rather than lying flat, patients are seated for scans. Now on sale in Japan, the system has received U.S. Food and Drug Administration approval. Patients robbed of muscle control by neurodegenerative disease can communicate via Yamamoto’s gear, which can sense the brain’s “yes” and “no” responses to questions. He hopes his device will further understanding of language processes. “We don’t know so much about the brain,” he says. “I would like to find new functional areas.”