This Article Full Text (PDF) Submit a scholarly reply Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Doyle, D. J. Search for Related Content PubMed Articles by Doyle, D. J.

Speech Recognition Comes of Age

Toronto, Ontario

I can remember when I was a student in my early university days, years before the advent of microprocessors and personal computers, wondering whether or not I would live long enough to witness speech recognition software that would work. I dreamed of software that would allow one to dictate directly to a computer. (A victim of a traditional college preparatory school thinking, I had neither chance nor motivation to learn to type in high school. I still type painfully using only two fingers.)

I took my first plunge into speech recognition technology about five years ago, buying a package offered by IBM that would only operate under OS/2. After an enormous amount of technical effort on the part of both myself and IBM we got the system to work on a 486 laptop but the performance was so poor it was still much easier to type with two fingers. About two years later, I bought an upgrade to the product and ran it on a faster computer, only to find that the performance was still unacceptable. More recently, I tried once again, this time getting the latest release of Dragon Naturally Speaking, Version 4, and ran it on an even faster computer. Once again, there were severe performance problems, mostly related to the volume of word inaccuracies. I was ready to give up.

Then it occurred to me that I might get better performance if I used a USB microphone instead of using an ordinary analog microphone plugged into the computer sound card. Apparently, there is a lot of electrical noise inside most computers, so that analog signals can become heavily contaminated before being processed. The problem is even worse with laptop computers, with their crowded electronics. However, this problem can be eliminated (at least on latest generation computers) by using a USB microphone. (USB stands for "universal serial bus" and is a new interface technology supported by Windows 98 that can be used for hooking up peripheral components to computers. USB technology can be used for printers, scanners, microphones, TV cameras, and even keyboards, mice and modems.)

With the addition of a USB microphone from Telex, a remarkable transformation occurred to my system. The first thing I noticed, after installing the USB microphone, was that vocabulary training took only a few minutes, and not the better part of an hour that I was accustomed to. Secondly, the performance of the system was finally good enough to use for real world dictation tasks. The accuracy of my system is now good enough to dictate real letters and write real reports, even with the noise of the hard drive and fan in the background. (In fact I can even get reasonable results with the TV playing gently in the background).

Yes, speech recognition has arrived and is finally practical. It is still imperfect, but it is definitely now worth while to two-finger typists. I can see that it will not be long before speech recognition technology will be available for clinical data entry during routine anesthesia care. Still, it is not clear that this technology will be sufficiently reliable in the noisy operating room environment that it will see widespread adoption for use in charting and documentation.

This Article Full Text (PDF) Submit a scholarly reply Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Doyle, D. J. Search for Related Content PubMed Articles by Doyle, D. J.