Voice Recognition Software

From "Ask Bob Rankin Archives" (a newsletter I subscribe to) Didja ever notice that Captain Kirk never had to fuss with a mouse or keyboard to communicate with the computer on Star Trek? He simply spoke direct commands at the machine, and the computer complied. Happily, voice recognition technology has moved from the bridge of the Enterprise to your home computer. Here's how to get started with voice computing.

 

What Is Voice Recognition Software?

Simply put, voice recognition software allows you to speak commands to a computer rather than use a mouse or keyboard. Though it may seem like a futuristic notion, voice computing or speech recognition has been around for over a decade, and is still a rapidly evolving technology. Even though it's not used by many on their home PCs, you probably have already encountered this type of data input. A lot of companies, especially telecom providers, are already using voice-recognition when you call their customer service centers. You can answer "Yes" or "No" or other specific phrases to questions that the voice-activated system asks, rather than pressing "1" for "Yes" or "2" for "No". Also, more and more vehicles are now coming equipped with voice recognition computers that let the driver operate the radio, CD player, or a cell phone.

How Does Voice Recognition Work?

Without getting too terribly geeky, a person's voice is translated into a digital format that a computer can read. Sound travels in waves, and the range of sounds we can hear are created by changing the height (amplitude) and frequency of those waves. So by examining the "voiceprint" of a spoken word or phrase, and comparing it to the sound waves created by speaking specific utterances, the computer can make a pretty good guess at what you said. In a nutshell, it's pattern matching.

In situations where the only options are "Yes" and "No" it's a lot easier for a computer to figure out which word you said, because they sound quite different. It obviously gets trickier when the voice recognition software has to figure out what you said in a larger context. And because human vocal patterns are as varied as humans, with individuals having different accents and inflections, voice recognition software must typically be trained to identify the way words are spoken by a particular user.

Want to Try Voice Recognition Software?

You can buy software available that will allow you to speak commands to your PC. Perhaps the most common reason to use voice recognition, with all apologies to Mavis Beacon, is to eliminate the need for typing by dictating. No more two-finger pecking at a keyboard to create letters, e-mails, school reports, and such; just sit back and talk at your PC. And of course it's a boon to those with physical disabilities.

The most popular voice recognition software currently available to accomplish this is Dragon Naturally Speaking, created by a company called Nuance. With this software, you can dictate text in a word processor such as MS Word, navigate through the Windows desktop or surf with Internet Explorer. The latest version, 9.0 Professional comes with a microphone, and the program comes with support for Bluetooth wireless headsets.
 
The installation is painless; simply follow the prompts. The most time-consuming part of using Dragon is training the program to recognize your voice. It can take about two weeks of consistent use to get the program to fully understand your words. But hey, that's still less time than it takes to get carpal tunnel from typing. Nuance boasts 95% accuracy with word recognition, but many users claim that the reality is more like 80-85%. There also seems to be an issue with dictating numbers in a spreadsheet program. If you are, for example, using Excel, and you say "10", Dragon will type out the word "ten". You have to say, "numeral 10." Be warned, you will also need at least a 1Ghz processor, 512 MB of RAM, 1 GB of free hard disk space and Windows XP/2000 to run this program efficiently.

Currently, Dragon Naturally Speaking is not available for Mac OS systems. The premiere Mac program for voice recognition is iListen, but you can try Apple's Speech Recognition feature built into OS X. Windows Vista also comes with its own voice recognition software, though some say it is not as accurate as Dragon's program. There are other programs available that will allow you to dictate to your computer, some are task-specific; only recognizing math functions or e-mail commands. Here's a comprehensive (but dated) list of voice recognition software options.

At this time, there is no one program that you can use "out-of-the-box" that will transcribe your dictation or your PC commands perfectly. Dragon is the leader of the pack in the PC market so far, but voice recognition has a way to go before perfectly recognizing words other than "Yes" or "No". So Captain Kirk, you can keep that smug look on your face, but watch your back... software developers are putting a lot of time and energy into creating better voice recognition programs.