Speech technology powers Instinct phone functions, other wireless devices

Voice or text input? Both likely to prevail for a long time

Do you talk to your cell phone?

Sure, we all talk into our cell phones, but do you talk to it? In other words, do you use speech to command the phone to find a number and make a call or conduct a search, instead of using your fingers to touch keys and type in numbers or text?

The new Instinct smart phone, which goes on sale Friday, has a number of speech-activation functions for calling, texting and finding information on traffic, movies, sports and weather.

Sprint Nextel Corp., which developed the Instinct with Samsung Electronics Co., has focused on the speech-command functionality as a clear differentiator for the Instinct, especially when compared with Apple Inc.'s iPhone. The iPhone 3G, another touch-screen smart phone, so far has no native speech commands. It ships on July 11.

Nuance Communications Inc. makes the speech-activation software used in the Instinct. Last week, the company announced that it has developed a prototype for voice-searching capability on the iPhone. However, Apple officials have yet to decide whether to endorse the Nuance product for the iPhone. It could be provided through the Apple App Store when it launches in July, according to Michael Wehrs, vice president of evangelism at Nuance in Burlington, Mass.

Whether Apple even wants speech commands in its iPhone 3G is unclear, since speech-command technology sometimes goes unused, one analyst noted.

Voice-activated calling is already widely available in some wireless devices, and Nuance has shipped its software on 200 million wireless BlackBerry devices from Research In Motion Ltd., as well as devices from Palm Inc., Wehrs said. Speech-recognition capabilities have been growing in cell phones for many years and account for much of the $200 million in annual revenue in Nuance's mobile division.

But Jack Gold, an analyst at J.Gold Associates, noted that "some users absolutely love voice commands, while some say it's not a big deal." He said the enactment of hands-free driving laws will clearly spark interest in voice-command software for cell phone users, although Microsoft Sync and other software in some vehicles have partly answered that need.

Still, there will always be the problem of making voice-activation software work well when there is background noise, such as while driving or in an airport, he said.

Voice-activation software "is not one of those killer apps that will sell a lot of phones, but a buyer might buy a phone and notice when it's missing," Gold added. While providing voice commands "would be a good addition to the iPhone," he also said the cell phone culture is many years away from Star Trek's promise of using voice commands almost exclusively to control a ubiquitous computer running in the background.

The Instinct has a voice-command functionality not seen on any other wireless device that allows a user to call up an address for a text or picture message via voice, Wehrs said. Once the application is launched, however, the user must still type in the message or find the picture to be sent through touch commands.

Nuance's speech-recognition software works by requiring a user to train a device for a few minutes so that it recognizes a person's voice, or the software can learn as a person makes commands and will correct mistakes as it goes, Wehrs said.

The speech-recognition software resides on the phone in its current form, but a next generation of Nuance software, shipping later in the year, will allow a cell phone to search wirelessly for a server operated by Nuance for longer speech dictation, Wehrs said. The server will relieve the phone from processing the information to reduce the drain on memory and battery.

Nuance aims for 99% accuracy in its software, but that figure can drop if the software determines that fixing every error might take too long, which would take up processor and battery output, Wehrs said.

Nuance also provides predictive text technology, which was shipped on 750 million phones globally last year -- about three-fourths of the entire market, Wehrs said. The technology, which is available in as many as 70 languages, helps reduce errors caused by so-called fat-fingering on a touch screen or on the phone's physical keys, he said. The technology is also included on the Instinct.

Wehrs predicted that speech and text input will continue in cell phones for a long time. Nuance's largest mobile customer, Nokia Corp., has said that it plans to provide a diverse array of input technologies in its iPhone-like device, code-named Tube, when it appears later this year, Wehrs said. He would not disclose any further details, but added, "Nokia wants to have the broadest range of options available."

Copyright © 2008 IDG Communications, Inc.

It’s time to break the ChatGPT habit
Shop Tech Products at Amazon