The Wireless World Will Be Voice-Activated

The technology is getting better, but it's not ready yet

Small form-factor wireless phones and personal digital assistants have pushed users about as far as they can go. Typing commands on a dial pad or using a mini-stylus to navigate a shrunken computer screen is appealing to only so many people. Users are fed up and want their voices heard — literally.

Many industry executives are now looking toward voice-recognition technology to solve the interface problem for wireless devices. But while computers are getting better at understanding the human voice, we are several years away from widespread deployment of voice-activated interfaces in wireless phones and PDAs.

"Vendors are making all the right noises, but, because of the complexities of initiating the technology, its not ready," said Rachel MacAulay, a Kelsey Group senior analyst. "However, we have complete faith theyll work out the problems."

But dont blame vendors for the delay. Bandwidth constraints, as well as the lack of supporting technologies, such as voice-enabled Web sites, will slow use of this promising interface.

Most industry experts say the technology can recognize voice commands with up to 95 percent accuracy, though there are plenty of caveats in that percentage.

Voice recognition is most accurate when users speak in a standard American dialect or have a mild accent. Also, most applications only work with limited vocabulary, perhaps only a few hundred words.

On the other hand, some vendors have proven the concept of "speaker independent" or "natural language" voice recognition, which allow users to speak in everyday sentences and be understood without "training" their devices.

The auto industry is proving one of the biggest boosters of voice recognition. Manufacturers are rolling out on-board computers for the car. The devices provide everything from emergency assistance services to Web-type info such as stock quotes, sports scores and weather. Because of safety concerns, auto manufacturers have designed these devices to operate solely with a voice interface.

Mike Peterson, a director of General Motors OnStar subsidiary, said voice recognition technology proved highly dependable during extensive tests last year. The voice-activated Virtual Advisor Internet portal now is available in 32 GM models.

OnStar is using products from Nuance Communications for server-based voice-recognition processing, as well as noise-filtering technology from Delphi Automotive Systems on the device side, and a speech-to-text application from SpeechWorks International.

There are still a few glitches, but as Peterson pointed out, GM would not be installing the Virtual Advisor if voice technology was still more press release than performance.

"We think it works very well right now," he said.