IBM-SPEECH RECOGNITION-HANDHELDS AND AUTOS
Almaden director seeks a 'delightful' revolution
By Charles J. Murray
EE Times
(01/24/01, 9:26 p.m. EST)
The work surfaces in Robert Morris' San Jose, Calif., office are cluttered with computing devices: two desktop PCs and four notebooks, as well as a pager, a cell phone, a pair of Think Pads and a Palm computer that Morris carries when he travels.
The director of IBM's Almaden Research Center is quick to show visitors that all six of his computers are connected: When one machine receives an e-mail, the other five quickly chime in, letting him know that they, too, have gotten the message.
"Pervasiveness means that when you walk into a room, you're walking into a computing environment," Morris said. "The user should have an experience — it should be easy to obtain information and easy to participate in that environment."
This glut of computing devices is proof that Morris practices what he preaches. As vice president of personal systems for IBM Corp. and the holder of a PhD in computer science from UCLA, Morris is a firm believer in the future of pervasive computing.
Intertwined with pervasiveness is ease of use. Morris' definition of "easy" hinges on the concept of a human-computer interface so transparent that users barely know it's there. The company is working on a variety of such interfaces, including so-called "attentive" techniques that watch a user's eye movements and facial expressions, as well as autonomous systems that communicate with one another via Bluetooth, IEEE-802.11 or infrared methods.
IBM engineers around the globe are also stepping up efforts to bring speech recognition and handwriting recognition to the fore, for use in handheld devices and automobiles. Morris believes that such efforts will serve as the key to acceptance of pervasive computing.
"To start a revolution in computing, you've got to have an ease-of-use breakthrough," Morris noted. "If you don't have that, your new technology is not a revolution. It's just a pain."
Indeed, Morris insists that all of computing's major revolutions were prompted by breakthroughs in ease of use. Batch computing, for example, was displaced because time-share computing was simpler, he said. Similarly, personal computers replaced time-share because of ease of use, and browsers brought the Internet to millions for the same reason.
"Ease of use is more fundamental than engineers give it credit for," he said.
Morris and other IBM executives don't pretend to know how the shakeout of today's computer products will affect current human-machine interfaces, such as the mouse and keyboard. But he foresees two potential paths for future computing devices. The first path results in a single winner — a universal device, such as a multifunction Internet phone — that does it all. The second involves a variety of devices — handhelds, desktop units, laptops, tablet PCs. In that scenario, Morris believes, all devices coexist, and all are capable of talking to one another.
In either case, the company's engineers are preparing for a future of simpler interfaces. At the Almaden facility, they have already developed an alternative to the "Qwerty" keyboard (so named because Q-W-E-R-T-Y compose the top line of letters on the left-hand side of all conventional keyboards). IBM's new technique is designed to address the problems of typing with a stylus on a tiny, handheld, virtual keyboard. The solution consists of a collection of hexagonal keys, with the most commonly used combinations of letters located in the middle, around the space key. Less commonly used letters are placed farther from the center.
IBM engineers say that their new keyboard reduces the distance a user's hand travels between letters and therefore speeds typing time on cellular keyboards by about 50 percent.
"This design is not optimized for two hands," said Barton Smith, a research-staff member at IBM Almaden. "It's designed for one tap."
Engineers at the Almaden facility are also working on a potential future in which the human-machine interface will involve intelligence. In an internal project known as Blue Eyes, they've created so-called "attentive computers" that use cameras to watch a user's eyes and facial expressions, and then infer his or her desired actions.
Attentive computing may serve as the ultimate user interface because it enables a machine to make decisions and take action on its own. An attentive computer, for example, might recognize that a user is in an important conference, and hold all e-mail messages until that conference has ended. Or it could make a decision to interrupt the user and deliver an urgent phone call.
Ultimately, such techniques hold promise for the disabled, who could access a computer interface through eye movements. Attentive computing could also serve as an aid to effectiveness in so-called "smart houses," which up to now have struggled with voice commands.
"Part of the problem with smart houses is that they don't 'know' when you're talking to them," said David Koons, a research engineer at IBM Almaden. "The vocabulary is often too broad for the system to understand. But by using attentive computing, you can narrow the context of your commands, and the computer has a better chance of understanding you."
Engineers say that the same concepts can also be applied to such devices as Internet-enabled telephones. "There's no reason your cell phone can't 'know' that you're sitting in an opera house and that it's supposed to be quiet," Smith said.
If the future holds a variety of devices, however, IBM engineers know they will need to put more development effort into such technologies as voice and handwriting recognition. Those systems enable computing companies to meet diverse needs for diverse users, ranging from rush-hour drivers to customers in cultures where pictographic languages are the norm.
IBM engineers are working on those technologies at the company's Watson Research Center in Yorktown Heights, N.Y., as well as in its research labs in New Delhi, India, and Beijing. "The pen [interface] is even more important in China and India than it is here," Morris said. "Today, there's not a good keyboard for those languages."
Although handwriting-recognition technology may still be years away from optimal performance, Morris said that pen interfaces are here today. Users can now employ pen technology to capture their own handwriting and store it as such, without the need for software to transcribe it into typed copy. "If we can electronically capture the ink, then we won't need handwriting-recognition technology right now," Morris said.
But speech has technological hurdles to cross before it can qualify as a revolution, he said. Today's best speech systems typically operate at accuracy levels slightly above 90 percent, said Morris, and they won't improve until engineers make breakthroughs in system intelligence.
"To get much higher than 90 percent accuracy, you actually have to understand the speech, not just the words," he said. "So the only way we're ever going to achieve 100 percent accuracy is by building domain knowledge."
Such breakthroughs could eventually constitute the missing ingredient to bring pervasive computing into the mainstream, Morris said. "In the end, we have to make the computing experience natural," he said. "Also, over and above being natural, it has to be delightful. Ultimately, the only way you'll start a revolution is to create a delightful computing experience.
