- BACK TO HELP
FIRST Unzip: Development : Use VAndLDvlp.vbp ,make
'VoiceAction.exe' !
Right now you MUST be
having with you the words you want VoiceAction to
recognize.
Pick your first word ...
Click 'Open' in 'Language Settings' , set ' Enable
Artificial Intelligence' to NO.
In the Basic Vocabulary Settings, set Number of
files to scan = 1.
In "Consider this File to Match No.1 to
5" , Click and enter NEW NAME with PATH). Also Enter
a NAME and PATH for "Consider this file to Add"
(This is a MUST). Of course at this stage the names of
Add and Match file no.1 have to be same. Remember to
enter all the 5 names but
Number of files to scan
has to be kept 1 since we are preparing the first one.
Now ' SAVE' these settings.
Set Sample Processing Resolution to
"COMPLETE" if you are recording a vyanjana or
to 4000 ( less) if you are unable to devise vyanjanas of
your word. But remember in this case to add only uncut
playback.
Now press "Add" and then
"Start".
After you hear ::start:: record your first WORD or
VYANJANA. Add the code you had already decided or as you
hear it when it asks Add Playback to File? Do not add
noise or unwanted stuff. *

After pressing
Y
es,
VoiceAction will display a input box
Enter the word you uttered or as you hear the
playback. (Above in pic, we were preparing a speech based
menu operations and are seen recording the mostly used
word ::File::)
For Sample Processing Resolution =
"COMPLETE" (or 0) it will playback the complete
word.
For other values of the sample processing
resolution, you will hear the playback for the specified
length and will be stored for that length in the file
only. This kind of discrete processing is advisable for
better recognition.
Prepare all your words in this first file. On doing
so STOP Recording. Enter the name given to Matching file
no. 2 in ; "Consider this file to Add".
Set Number of files to scan to 2.
REPEAT THE ABOVE RECORDING FOR YOUR FIRST WORD FOR
ALL THE FOUR FILES.
(Note that you can prepare 10
such files)
Input absolutely the same Characters which you had
entered for the first file for all the files. (Note that
you can prepare 10 such files, in other words, it is a
record of 10 instances of your particular word)
At this stage now, you may pick your second word
and repeat the above process for that particular word
OR
if you wish to test your recognition for first word then
move to the next step below.
AFTER THIS ...
PRESS ' Don't ADD.'
Keep Sample Processing Resolution to the same value
which was used during above recordings.
Say the word again.(i.e. the one you had added)
Now here we have recorded in the File
"Tuto1.SAY", menu required words, like ; File,
Edit, View
When we say ::File:: which was added as
shown above, the percentage match shows ::File:: has
achieved 67% and is confirmed.
- FOR Sample Processing Resolution =
"COMPLETE"
- FOR Sample Processing Resolution = 1000
This is a different file in which we had stored
divisons of ::File:: (/1000)
Press "Start" again and see the bottom
left box strings. If they are not produced as you wish,
then you probably haven't entered a common code for every
vyanjana.
USE THE SAME CHARACTER CODES FOR EVERY FILE, THIS
IS CRUCIAL. USE VOCABULARY EDITOR ON EVERY FILE TO EDIT
WHAT YOU ENTER.
TIP : See which of the characters get replaced by
each other. See if they get pronounced in the similar or
nearly the similar way, neutralize them by giving a new
common name or use one of their titles on every one of
them. As example ai / ei/ ii = ii.
THIS IS THE TIME TO EXPERIMENT WITH Accuracy's,
Sample processing Resolution.
- Now lets see Artificial
Intelligence
Now in Language Settings set "Enable
Artificial Intelligence" and "Add to
Language" to YES. Also fill up the new file names
for your 'Language files to Add' and 'Language files to
Match'.
Press 'Start' and say your first word again.
This
setting displays the 'Language Developer'. To see the
Golden Character that is produced for the current word,
go to menu . . . Current Memory....Display Memory.
Enter your word in your font setting and also enter
the rating ( in comparison to the ideal ) . Press
"Add" to input this data. And then press
"Add" (of the Drive Icon) to save this in the
specified Language file to Add.
REPEAT THIS FOR ALL
YOUR WORDS.
WHEN YOU ARE ADDING FOR THE SAME WORD BE SURE TO PRESS
"RESAVE".
REPEAT EVERY WORD ATLEAST 10
TIMES.(To design a final protocol )
SEE THE ON
SCREEN AVAILABLE PROTOCOL DESIGN EXAMPLES AND DESIGN A
PROTOCOL FOR THE RESPONSE FOR EVERY PARTICULAR WORD AND
ENTER IT WITH RATING!
Note that
rating plays a important role in final output.
DISABLE ALL THE DEVEOPMENT RELATED SETTINGS LIKE
UNRECOGNIZED NOTIFICATIONS AND ADDITIONS.
YOUR WORK IS OVER, USE THE PLATINUM WORD 1 (and its
other brothers) WHEREVER AND FOR WHATEVER YOU DESIRE !
Thank you for
using VoiceAction Component !
*{ If you think that your current
playback has truncated words then...
Prepare vyanjanas of that particular
patch ... As a example .."eight"
Use your common sense to divide
this, it gets divided in "A" +" th "
Prepare a list of vyanjanas for all
your words, you will find a lot of
common things. Make the list as
short as possible. For the navigation
thing i.e. right, left, up , down we
prepared
a ( like in up)
aa ( like in ask)
un ( like in down)
th ( like in eight)
A.( like in eight)
Record these as they sound.
See Examples : Basic Recognition }
Remember
the design key :
Control over
every word, through wav file playback announcement.
Program your
dialogues
with the user through proper scripting. Watchout for the
coming upgradations of VoiceAction with such dialogues
scripting !
BACK TO HELP