Voice Box New Products Recognize, Respond to Human Speech

Article excerpt

When Captain Kirk issued verbal commands to the computer aboard the U.S.S. Enterprise in the old "Star Trek" TV series, it seemed impossible that anyone could really get a machine to respond to the human voice.

Nearly 30 years later, life has surpassed art. Computer and software makers including IBM and Microsoft Corp. are marketing speech recognition products that allow computers to operate by voice. By simply speaking into a microphone, a user can execute a computer's functions and enter text without touching a keyboard.

"This `hands free' idea of speech recognition is starting to grow," said Nancy Jamison, a computer industry analyst for Dataquest.

There are two types of speech recognition systems. One is called a voice navigator, an inexpensive software system that puts the user in control of the computer's commands, such as opening files and moving text.

Dictation systems, which run about $1,000, enter text while a user talks. They also manage the computer's applications.

Until recently, speech recognition systems were found in expensive specialty computers, such as those for the handicapped. But in the last year, the technology has moved toward more broad-based applications.

Many professionals who depend on dictation, like lawyers, doctors and journalists, have taken to these systems, as have the disabled.

Dr. Stephen Herman, a radiologist at The Toronto Hospital, uses the IBM VoiceType Dictation system for OS-2 to do most of his lab and research reports. What used to take days to put together, he accomplishes in one sitting.

"I can dictate my own report, edit it and send it directly to the doctor who needs it," Herman said. "It saves time and steps."

IBM launched VoiceType in the fall of 1993. It's capable of recognizing 32,000 spoken words at approximately 70-100 words a minute, with about 97 percent accuracy. A two-hour enrollment session familiarizes the computer with a user's voice. Once this session is completed, the system is ready for use.

Like most speech systems, VoiceType, which is also available for Windows, works by identifying the speaker's sound and language pattern, and then converting those phonemes into text. A statistical model breaks down the speech and creates a probable word sequence.

DragonDictate for Windows and Kurzweil Voice for Windows are other popular dictation systems. Both have vast vocabularies that can be updated and require enrollment sessions, which run under 30 minutes. …