Continuous Speech Recognition:  Opening Windows of Opportunity

Alexia Jones

Continuous speech recognition software, often referred to as voice recognition software, makes it possible for computers to decode human speech. With this software, people can control computers by dictating commands, and can compose email and word processing documents by using dictation. Speech recognition software allows the user to generate text by talking into a microphone instead of typing. 

When first introduced, speech recognition software had very few and very specialized applications. The applications helped people with disabilities or repetitive stress injuries. Early versions relied on discrete speech (halted, word-by-word fashion). Advancements in technology increased the program’s usefulness by allowing for natural or continuous speech. Technology also allowed the applications to include functionality in fields such as medicine, law, and education. 

Dragon Naturally Speaking, one of the most well-known speech recognition programs, works by using simple voice commands with a special microphone that comes packaged with the software. These commands transfer directly to the computer screen. Dragon, compatible with most Windows systems, requires only a sound card and microphone, significantly increasing the speed of producing text. 

For example, most people type an average of fifty words per minute. Therefore, a 900-word document takes approximately eighteen minutes to type. Using continuous speech recognition software, a user can speak 140 to 160 words per minute, which reduces eighteen minutes of typing to a mere 6.5 minutes. 

The software for Dragon requires an extended training session during which the computer system learns to recognize the user’s voice. The user accomplishes this by spending twenty to thirty minutes reading passages from a novel. Users can later train the system to recognize words not included in its vocabulary. Drawbacks exist, however. The computer substitutes a real word for a word it does not understand. Users must carefully proofread documents for such linguistic errors. They must also maintain checks for correct usage of homonyms such as there, their, and they’re. 

With speech recognition software, windows of opportunity, which did not exist a few short years ago, open wide with possibilities for people with both physical and learning disabilities. 

As prices decrease and performance improves, speech recognition software enters the mainstream as a viable alternative to keyboards.

Editors | Current Issue | About STC | Links | Disclaimer