THIS CODE RUNS/COMPILES ON 5.1 Irix OR LATER
THIS CODE WILL NOT RUN ON LESS THAN 5.1 Irix
THIS DEMO IS BUILT WITH PRE-RELEASED DIGITAL MEDIA LIBRARY CODE.
THE API AND FUNCTIONALITY ARE SUBJECT TO CHANGE. THE FINAL
RELEASED VERSIONS OF THE DIGITAL MEDIA LIBRARIES WILL BE
AVAILABLE AT THE END OF THIS YEAR BY ORDERING THE DIGITAL MEDIA
LIBRARY DEVELOPMENT OPTION, "SC4-DEMDEV-1.2".
THE SPEECH RECOGNITION DEVELOPER TOOLKIT IS AVAILABLE FROM:
SCOTT INSTRUMENTS CORP.
1111 WILLOW SPRINGS DRIVE
DENTON, TX 76205
TEL (817) 387-9514
FAX (817) 566-3174
______________________________________________________________________________
~4Dgifts/toolbox/src/exampleCode/speech/lackey README
mags 04.04.94
lackey
This is a speech recognition application example. It recognizes
speech through the use of a speech recognition library. The
example uses speech to launch desktop applications. Lackey has an
internal list of "words" (words to be recognized) and their
corresponding applications to be launched. Saying one of these
words causes the execution of that application . The current set
of commands recognize by lackey are clock, shell, apanel, and
lackey, which can easily be extended.
An audio capable system (Indigo, Indigo2, Indy) and a microphone are
required.
This example program was developed using SGI's speech recognition
software. This is an introduction to using the speech C++ API and
the architecture of the speech software system.
INSTALLATION:
Before you can begin using this application you must update your
system software to include the speech components. Included on this
edition of the Developers Toolbox are the inst modules for the speech
execution and developer subsets. When installed will update your
system with the speech DSO for your Xserver, the speech templates for
existing applications (such as Showcase), sounds and images, the speech
client library, include files, and sample programs.
The speech server is part of the Xserver. Therefore, to activate
the speech recognition restart your Xserver by logging out of
your current session.
After you are logged back into your active session, start the
Speech Recognition panel :
% srpanel &
STARTING lackey:
Be sure to set your apanel settings as follows:
Input sampling rate at 8khz, the input device to microphone and
the input level to 10.
Now you can start lackey
% lackey &
TRAINING THE WORDS:
None of the words that lackey will react to are known to the speech system
You mustyou must train the recognizer to recognize them. Because this is
a speaker independent system, the more different people that train the
words, the better the recognizer will get at recognizing variances in
the different speakers that use the system. Since you have
never trained the words to be recognized by the lackey program
the speech recognition panel will prompt you to train each of
these new words one at a time. This will be denoted by a picture
of a cat in the image window of the srpanel. Make sure the
microphone at least 12 inches away from you. Slowly repeat the
word (displayed prompt window of srpanel) at a normal tone until the
word is recognized (at a minimum this will take 4 samples). You will see
(1/4) near the word being trained. This represents that one of four
valid samples have collected. Keep repeating the word until all four
samples have been collected. Repeat this process for the rest of the
words in lackeys vocabulary. If this process some how aborts or fails,
you can use the Customization panel found in the srpanel's
Recognizer pulldown menu.
Features of SGI's upcoming release of speech technology:
- speaker-independent discrete-utterance recognition
- quick response (less than 200 msecs)
- medium-sized vocabularies (50 active words at a time)
- no extra hardware required
- server-based
- supports multiple speech application clients
- supports networked speech application clients
- handles focus policies for speech application clients
- dispatches recognition events
- audio samples processed only once by central server
(for computational efficiency)
- pretrained words and phrases
- a suite of selected applications that are "speech-aware"
- CASE tools
- Showcase
- Desktop
- a vocabulary development system that allows the user to add,
modify, and delete the words which can be recognized
- a control panel through which the user can set behavioral
characteristics such as the acceptance and rejection thresholds
- a tool to generate actions for applications that are not
"speech aware" (not listening for speech input)
- documentation
SGI's developers product for speech-applications also feature:
- and API implemented with C and C++ speech headers & libraries
- advanced vocabulary development tools
- a database of pretrained words and phrases
- documentation
- programmers' guide
- API specification
- policy (style) guide for developing speech application
behavior and vocabularies
Source
Documentation
Audio
Reference
OR, ...