Hi,
Although there are a few systems providing speech to text, the
fundamental process, that is speech recognition, is still at a stage
when the science-fiction stuff you see in the movies actually becomes
a fully viable reality!
Another issue with what you want to do is that you have your speech
recorded into micro cassette. Almost all speech recognition systems
need 'crystal-clear' sound because the technology is still at a stage
when it cannot ignore too much background noise and intonation
problems (not with the way you speak but with the way your speech is
reproduced from this media!). In fact these systems maintain that you
will need a very good microphone (they certify certain products
against others that will work with their systems) and a noise-free
environment.
Yet another issue is that most speech recognition software requires to
be 'trained' with a particular voice before it begins to recognize
words correctly. I have personally tried using the Microsoft Windows
XP speech recognition system and have come up with hilarious results!
It does have to do with my accent, but then any human voice, including
yours has an accent. After all, we cannot talk like those speaking
machines you hear in the movies with synthesized voices! Not to
mention the overall vocabulary of a doll that says "I'm hungry!"
However Apple claims its software does not need any training. (The URL
is listed below.)
I do feel that what you want to do is a tall order. One, because of
the nature of what you want to do and two, because the software
technology is not yet geared up for this.
If you want to try out any of the software options listed below, you
will have to do one thing first. You will have to transfer all your
cassettes to the hard drive and maybe spend some time 'cleaning it up'
before you can apply the speech recognition software. If you do want
to go ahead and try it out, I can run you through how you have to do
this 'clean-up' procedure. Of course, this is as big a question as
this one!
-----ISSUES---------
From PC MAGAZINE
-URL-
http://www.pcmag.com/article2/0,4149,479327,00.asp
Text to Speech to Text
I conduct interviews, which I record. Is there a program that can
translate these interviews into editable text?
In theory, yes. Your only hope is to purchase a speech recognition
system like IBM's ViaVoice or ScanSoft's Dragon NaturallySpeaking and
try to feed it audio from your handheld voice recorder. There's a
slight chance it will be clever enough to turn your interviews into
coherent text.
Whether you plug your recorder directly into your sound card or simply
hold it up to the PC's microphone, even the best speech recognition
apps will turn your recorded audio into gibberish. Trying both methods
with an analog Panasonic recorder and the latest version of ViaVoice,
we couldn't get a single word converted into text.
The problem is that a program like ViaVoice doesn't work well unless
it's been trained to understand a specific person's voice and to
compensate for a particular environment's ambient noise.
Cade Metz, senior writer
Copyright © 2002 Ziff Davis Media Inc. All Rights Reserved.
----------
------- SOME RESOURCES -------
Scansoft, a company that makes Dragon Speech recognition software has
an opinion about speech recognition... See White Papers from
www.scansoft.com (requires registration)
-URL-
http://www.scansoft.com/whitepapers/
How Dragon NaturallySpeaking works
-URL-
http://www.talktoyourcomputer.com/howdnsworks.htm
The Speech Recognition Forum
-URL-
http://www.voicerecognition.com/forums
From Your Lips To Your Printer - article by James Fallows
-URL-
http://www.theatlantic.com/issues/2000/12/fallows.htm
Speech Recognition Technology: Maximizing Efficiency in Document
Preparation and Case Management
-URL-
http://www.ngtvoice.com/services/wstla.htm
-----------
-------- Some speech recognition programs -------
VIVA
-URL-
http://www.philipsspeechprocessing.com/
Dragon Naturally Speaking
-URL-
http://www.scansoft.com/naturallyspeaking/
Apple Speech Recognition
-URL-
http://www.apple.com/macos/speech/
IBM VIAVOICE for Windows and Mac
http://www-3.ibm.com/software/speech/
*** This is an open source ***
The CMU Sphinx Group Open Source Speech Recognition Engines for
GNU/Linux, UNIX variants, and Windows NT or later
-URL-
http://fife.speech.cs.cmu.edu/sphinx/
SpeechMagic from Phillips
-URL-
http://www.speech.philips.com/pc/pages/pc_home.htm
Microsoft Windows XP and Office 2002 have in-built speech recognition
software.
warm regards
vinods-ga |