|
|
|
Tips and tricks: Over to voice
IT MIGHT come as a surprise to you that the first paragraph of this article was written using a highly respected speech recognition program. Today almost any Windows PC can be programmed to recognize and respond to the user’s voice, converting speech into text, writing letters, composing emails or faxes.
It took a little over ten minutes to enter those fifty or so words, which is why the rest of this piece is being typed. In the past few years, speech recognition software has improved in leaps and bounds but even the very best systems still require a lot of patience and hard work to get them up to a useful speed.
In the early stages it’s not much faster than one-fingered, hunt-and-peck typing and it all goes to pot if you catch a cold, lose your voice or work in a noisy environment. However, what if you cannot type either through physical disability or dislike/fear of the keyboard? In that case, voice recognition is a practical way of getting your words into a PC.
But do not expect instant hands-free computing. Even the fastest and most efficient voice recognition programs have to go through a long-winded training process, to learn the user’s voice before they can achieve worthwhile accuracy.
Moreover some speech recognition systems make heavy weather of menu commands and program controls so you could still end up using the keyboard and/or mouse for a lot of routine operations.
Although several voice recognition programs are available that work on older and slower PCs, for best results you will need a fairly recent and speedy Pentium or Pentium-class multimedia machine.
Other basic requirements are at least Windows 95 or 98, 16Mb of RAM or more and more than 60 megabytes of free hard disc space. The PC should have an up-to-date soundcard, plus a microphone or headset with a boom mike.
A headset-type microphone is preferable as these are less sensitive to background noises and their performance is usually more consistent.
The three big players in the market are Dragon Systems, IBM and Talking Technologies. Microsoft is also taking a very keen interest in the technology. Its website boasts of a lot of information and some interesting downloads for developers and experimenters.
These include beta version of the voice recognition and text-to-speech ‘engines’ — a self-contained program designed to do a specific task that operates within a larger application — called SAPI (Speech Application Programming Interface).
Users of Microsoft Word might also like to try a program called Kurzweil Voice Commands, which controls all of the word processor’s functions using simple spoken phrases, like ‘select the next two paragraphs’ or ‘set this word to lowercase’. A fully working 60-day trial version is available from the Office Update section of the Microsoft website.
Installing a voice recognition program from CD-ROM normally only takes a few minutes, and that’s when the fun begins. The program’s first action is to go into a set-up routine, to measure background noise levels then test and adjust the sensitivity of the microphone. The amount of vocal training these packages require varies quite a lot.
Dragon Naturally Speaking, for example, depends on the user reading long passages of text (Arthur C. Clark’s 3001). The initial session can easily last more than half an hour, and it needs constant training to maintain accuracy. Kurzweil Voice, on the other hand, can be up and running in a few minutes, though first attempts are likely to be riddled with mistakes since it learns as it goes along.
The programs also work in slightly different ways. Most of them have the option (though not always included as standard) to enter text directly onto the page of your chosen word processor, as you speak. Others have their own text window.
Once a file has been created and corrected it is exported to an application, such as a word processor, fax program or internet e-mail window.
Even after a program has been fully trained it is important to remember to speak clearly and slowly using a consistently natural voice and accent, or at least the same voice that you used when setting up the program. Above all keep calm!
Even the best programs make frequent mistakes and there is a natural tendency to raise the pitch of the voice in frustration, which makes it even harder for the program to follow what you’re saying.
Voice recognition is a fast moving technology with huge vested interests in getting it to work properly. There is absolutely no doubt that within a very few years we’ll all be chatting merrily to our PCs and the keyboard will become obsolete.
The writer is a freelance contributor
|
|
|
|
