Fueled by increasing safety concerns and customer demand, the cellular market has become the fastest-growing segment for speech-recognition use in the United States, according to industry analysts.
Being able to command a cellular phone to “call office”-instead of requiring users to fumble for their handsets during rush hour stop-and-go, look down from the bumper of the car ahead and punch in a long string of numbers-is becoming an increasingly popular feature.
After years of uncertainty about the technology, new capabilities and higher accuracy rates have put speech recognition in the driver’s seat. Cellular carriers, including Ameritech Cellular Services, AT&T Wireless Services, Cellular One and SBC Communications Inc., now have deployed network-based voice-activated dialing systems in more than 40 cities across the country.
Cedar Knolls, N.J.-based Probe Research Inc. projected in a recent report that the entire telephony-based speech-recognition market will expand from less than $300 million in revenue in 1995 to more than $3.2 billion in 2000, with an annual average growth rate of 62 percent. This includes not only voice-dialing applications for cellular and wireline telephony, but also operator-services automation, enhanced-services automation and interactive television.
The report projects total revenue from voice-dialing features-for both wireline and wireless-will grow from $140 million in 1995 to $2.27 billion in 2000.
It also predicts that all cellular companies will offer voice dialing by 1997 and that 40 percent of all cellular subscribers will use the feature by the end of 2000.
“The cellular phone market offers perhaps the greatest opportunity for speech-recognition vendors,” stated the report. “Demand for VAD and voice access to enhanced services for cellular phones is growing due to the restrictive environments in which they are used.”
Voice recognition is providing cellular carriers with a key weapon in the battle of the cellular safety scare. The safety issue is a top priority for the cellular industry-it “scares the hell out of” cellular vendors and carriers, according to one source, and companies are anxious to support all technologies that offer a solution.
Cellular phone use while operating a vehicle without hands-free capability is illegal in many countries including Australia, Brazil and England and legislative pressure also is mounting here in the United States. Because of this, speech-recognition vendors believe voice dialing will be the future standard for cellular.
“I believe that [regulations] will take another two to three years, but I believe they are coming,” said Shaul Berger, vice president of sales and marketing for Cupertino, Calif.-based DSP Communications, a speech-recognition chip maker.
Safety concerns already have prompted regulators in several states-including California, Massachusetts and New York-to initiate legislation mandating hands-free-only use of cellular phones while driving.
No specific “hands-free” laws have been passed. However, legislators in Pennsylvania this summer passed a measure making it legal for cellular phone subscribers to use one-ear headsets while driving.
Competition
For cellular carriers, another key reason for offering voice-command features is competition.
With the PCS option coming soon to lure customers, cellular carriers are concerned about three issues to be competitive-cost, coverage and convenience, said Canby Dautel, director of sales for Cambridge, Mass.-based Voice Processing Corp., a major supplier of voice-dialing technology. The company sells the VProCel software for the cellular environment.
`They’re all going to get beaten up on cost,” he said. “And the way to win over that consumer marketplace is with additional features and passing on features such as voice-activated dialing.”
Although cellular may be getting a jump start on the voice-dialing game, the feature would likely be a natural target for PCS as well.
Ubiquitous voice dialing would allow PCS handset manufacturers, who will strive for a smaller pocket phone, to eliminate the keypad and therefore make the phones tinier and lighter, said Tom Aley, author of the Probe Research report.
Aley claims phone manufacturers already are working on this type of handset.
Technology advances
Running parallel with safety concerns, recent advances in speech recognition also are demanding that the cellular industry take notice. After years of skepticism over accuracy and time-to-market issues, voice recognition is viable now for voice dialing because large vocabularies are not needed, while new capabilities have come on-line that are key for the voice-dialing environment.
While there have been a lot of speech-recognition algorithms developed during the last 15 years, they’ve mainly targeted high vocabularies to serve the dictation market, said Berger. “That type of application still hasn’t reached the desired performance.”
But for applications like hands-free cellular operation, very few vocabulary words are needed, and that is one of the reasons why voice-dialing has been the first successful mass-market speech-recognition application. In addition, the cellular industry has been attracted to the following:
Speaker independence. In the past 18 months, rapid advances have been made toward speaker-independence, which allows multiple users to operate a single device without having to program individual voices and extensively train the system. This allows a husband and wife, for example, to both use the same cellular phone to give generic commands such as “send.” Speaker-dependent capabilities, however, still are important for interpreting personal directories programmed into the system that allow a user to instruct it to do more individualized tasks, such as command, “call Mary.” Most systems now incorporate both capabilities.
Continuous speech. Vendors now are touting systems capable of deciphering continuous speech, which allows users to speak in a natural tone and flow. This is an advance over “discrete speech” systems that require the user to speak each word or command separately and distinctly. It is particularly important for voice-dialing applications so that callers simply can say “call 1-303-555-1212” rather than pausing between each number or saying each number after a system-provided beep.
Accuracy. Voice recognition accuracy rates continue to rise. Vendors now are claiming rates between 95 percent and 99 percent for natural speech.
Noise/Music Cancellation. Many vendors already have added audio cancellation features to their products to specifically accommodate applications such as cellular phone use in the car, where the stereo or external noise may interfere.
Power/Cost. Driving all this are the continual leaps in processing power and declines in prices for chips.
Solutions
Most of the cellular voice-dialing systems now deployed are network-based applications, where the equipment is installed at the switch office and subscribers don’t need any special equipment.
The leading supplier of this technology is Dallas-based Voice Control Systems. VCS claims it holds 90 percent of the voice-dialing market for cellular carriers. Its chief entry into the cellular carrier market is through technology integrator Brite Voice Systems, Wichita, Kan., which offers the VoiceSelect system, using VCS technology.
Primarily because of market conditions, most voice-dialing is deployed in analog systems. However, VPC’s technology is being used for Southwestern Bell’s Time Division Multiple Access digital network in Chicago, and VCS also has its technology in TDMA networks, as well as in Hutchison Telephone’s Global System for Mobile communications network in Hong Kong. Neither company is working yet with potential Code Division Multiple Access carriers.
Digital networks theoretically would improve accuracy rates for network-based speech recognition capabilities because
the transmission should be clearer.
However, the TDMA environment surprisingly offers some disadvantages as well as advantages, according to Pete Foster, president and CEO of VCS.
“The advantages are that one has less noise of one type or another,” he explains. “With an analog channel you have the microphone open the whole time. The thing is just transmitting continuously so that … in between you speaking, noise is transmitted.
“With digital there’s less of that. The digital system works so that only when you’re speaking does it transmit, and the rest of the time it’s quieter. So that sounds better … and also makes the job of voice recognition easier.”
The problem, he said, is that there’s still a lot of RF kinks to iron out, which currently are making the TDMA channels noisier.
DSP Communications has a solution to that and other network-based speech recognition difficulties: deploy the technology at the end device instead. DSPC just introduced a new chip, with samples due out in the first quarter of 1996, that is designed to put the enabling technology in cellular phone adapter kits directly in the cars themselves-as well as for other applications such as car navigation systems. Users could have voice-dialing even where carriers don’t currently offer it.
The company’s new chip has achieved significantly high accuracy rates of 97 percent even in noisy environments such as cars, said Berger. The plus side, too, is that the system doesn’t have to deal with the additional noise that transmission over a cellular channel adds.
The other advantage to DSPC’s chip is that it offers speaker independent and speaker dependent capabilities on one chip, reducing costs.
Future advances
Continuing advances will bring new features to voice-dialing systems. VCS’s Foster said future versions of his company’s product will allow:
1) directory assistance call completion so customers don’t have to worry about dialing the number in the car;
2) speaker verification to prevent fraudulent cellular phone usage; and
3) fully voice-controlled voice mail access using words and not just numbers. Current technology allows users to access voice mail by saying, for example “voice mail” or “messages,” but subscribers don’t have much control of it unless they can memorize all of the numbers necessary to complete the functions. Future access will allow using the words “delete,” “save,” etc., to operate the system.
Dianne Hammer is a freelance telecommunications writer based in Denver, Colo.
