2
LAS VEGAS — “TV, I want to watch that Tom Cruise movie from the 1980s, you know, the one with all the fighter jets?”
One day that will be all a television needs to know to understand that a viewer wants to watch Top Gun.
Electronics companies are changing the way people interact with technology, creating new interfaces ranging from speech and visual recognition devices to touch screens and motion sensors that are taking human-computer relations to new realms.
This week in Las Vegas at the International Consumer Electronics show, the future was on display as everyone from Panasonic to Microsoft Corp. teased those in attendance with sneak peeks at the latest prototypes from their research labs.
These included wall-sized TVs that recognize who is standing in front of them, car stereo systems that can be operated by voice commands and visual recognition mobile devices that tell the user what movies are playing at a theatre simply by pointing it at the marquee. “We'll see things that are different, and novel interfaces that use computer vision to basically allow computers to see the environment like we humans can,” said Parham Aarabi, an engineering professor at the University of Toronto.
But don't toss out the television remote just yet – most of these technologies are still in the developmental phase and are years away from making it to the mass market.
However, the success of game-changing products like Apple Inc.'s touch-screen iPhone and Nintendo's motion-sensing Wii controllers are proof that the face of technology is changing.
The most prevalent of these technologies – speech recognition – isn't really all that new. Anyone who has ever reached an automated customer service agent on the other end of the phone knows this.
“People have built little applications to do that sort of thing for quite some time,” said Nelson Morgan, a professor of electrical engineering at the University of California at Berkeley.
According to Mr. Morgan, who is also the director of the International Computer Science Institute, there are three levels of speech recognition technology. The first recognizes simple commands; for instance, when a user says “five” to a television using this technology it would flip to channel five, a process akin to the one employed by telephone operator systems.
“Nobody wanted the first one,” he said. “Where speech recognition is useful is when it can do something that is not easily obtained with some other input mechanism.”
The second level of applications recognizes what the user wants by drawing meaning from continuous speech, for example when a user says “give me channel five,” which is significantly more complicated. Microsoft's in-car Sync software allows users to control the radio and their phones by using their voice and is one application of this technology. The third level is what researchers are aiming to perfect and involves the aforementioned Top Gun example.
However, unless users see an advantage to speech recognition technology over simply pressing a button, demand and adoption rates will be low, Mr. Morgan said.
“If you're replacing a bunch of menus … if you can have something that is flat, that you can say what you want and get it, then it's useful,” he said.
At this year's CES, in-car GPS systems from companies such as Garmin Ltd. were the most common new adopter of this kind of technology. However, initial reviews of some of these products was average at best.
Speech recognition systems still have many bugs to be worked out before all a user needs to say is “Top Gun” to bring up the movie, Mr. Aarabi said.
“We can get speech recognition systems that are very accurate – when you talk they understand the words very accurately,” he said. “The problem is that as soon as the noise environment changes – for example if you're in an office or a living room where there are kids or in a car – the recognition accuracy rates become very small, they drop.”
Currently, most speech recognition software would cause the television to change the channel by itself, if a user was watching Top Gun and Tom Cruise yelled out a number.
Although it may seem counterintuitive, the problems with visual recognition technology are actually easier to navigate, Mr. Aarabi said.
“We're not too far away from that right now, there are a variety of technologies where cameras on the street can recognize people,” he said. “The technology is there now; for better or worse that's something we might see within one or two decades.”
During his keynote address at CES, Microsoft chairman Bill Gates demonstrated software that can recognize people or places, and he predicted it will eventually be available on Windows Mobile devices. By pointing the device at a movie theatre for example, it would enable the user to buy tickets to the show and offer directions to nearby restaurants.
Technologies that can recognize people from afar and recognize and identify speech patterns come with heavy ethical implications. It isn't hard to imagine how that kind of scanning technology could be used for surveillance and security purposes.
