Voice Sencetive Screen Saver for NT4.0/2000/XP (v1.3)

1.Description:

This screen saver lets you unlock your workstation by
saying a key phrase (in addition to the standart way of doing it).

2.Howto

First of all, the program should learn your voice and a key phrase.
Select Display from the Control panel, then select "Screen saver" set 
"Voice Sensitive" and press the settings button. 

You will see a setup window where you can set an acceptance threshold 
using the slider on the left side of the window. This threshold
will be used to distinguish noise from sound to analyze. The slider
doesn't change settings of your sound card. You should use standard 
windows controls in order to enable recording from a microphone and 
adjust recording level.

The "Select phrase" selection box is there for future purposes, but
it can be useful anyway. You can type in your key phrase in 
scrsvr.ini file in your user profile directory,
and the text will be shown in the box as a reminder.

After setting the threshold, click "Train" button to proceed 
to the learning dialog box. The learning dialog box has three 
sliders:
 -the top one is used to set minimum phrase length
 -the idle slider sets the minimum period of silence indicating
  the end of a phrase being spoken
 -the activity slider sets minimal period of sound above the noise
  threshold indicating beginning of a phrase

Any sound sample longer than the minimum phrase length
is accepted for analisis. 

After the analisis you will see results as enlighted word "right" 
(the sample would be accepted) or "wrong" (the sample would be 
rejected). At this point you have to let the program know the correct
answer by pressing "Right" or "Wrong" buttons. The "Right" button
should be clicked if the phrase just recorded is your key phrase.
The "Wrong" button should be clicked only in case the program accepted
somebody else's voice or noise as a correct pasword. If the program
was correct about the phrase recorded or you feel like recording of your
password phrase was not clear and loud, click the "Skip" button. As
soon as you tell the program either it should learn the recorded sample 
as the right or wrong phrase, the sound sample will be placed in 
appropriate buffer for right or wrong samples (right samples buffer 
can save up to five samples). The program, then, will spent some time
learning the samples in the buffers. After that, it will switch back 
to the recording state.

Again, if you are not sure that the sample is recorded well, or you 
just don't want to use the sample for learning, you should click
the "Skip" button.

You have to press Ok button to save the results of
learning.

The "Reset" button resets program to initial learning state.
Click this button if you cannont succeed in teaching for a long
time (25-30 learning cycles).

After all, don't forget to set the "password protected" checkbox 
and the screen saver activation timeout.


3.Suggestions and useful information

3.1 Be prepared for a hard work. You will need to pronounce
your key phrase about 5-15 times, depending on your system 
performance and hardware. If the program is not able to 
accept your phrase with all-right quality after about 15 learning 
cicles try to press "Right" button, only in case it answers "wrong" 
on your key phrase, otherwise, if the key phrase is accepted
as right, press "Skip" button.

3.2 Your microphone should be tuned up very well. The progress bar in 
"Recording" dialog box should be moving only in case you are 
pronouncing a phrase not when you are waving the microphone or breathing
in it. Of course the progress bar is moving regardless of your 
efforts during the learning process.

3.3 Try not to move you microphone when you are speaking and short 
time before and after that. Dont exshale in it like you are trying 
to blow it away.

3.4 You had better to use 3 seconds length tongue twister
which you have known for a long time, your full name if it's 
long enough, or something like this as the key phrase. Don't force 
yourself to pronounce the phrase in special way. You should speak 
aloud, not forcing yourself, just speak as you always do.

3.5 Try to say the phrase one or two times before learning.

3.6 Don't use your voice samples as wrong ones.

3.7 Phrase buffers are not saved with learning results, so you need 
to pronounce five right samples to fill them in in case you reopen 
the train window.

3.8 Don't worry if the program shows "Learning failed" during
learning cycles.

3.9 The longer phrase your use the better security you have, but be 
careful, very long one can exceed phrase buffer length.

3.10 Be careful with MS GINA replacement. It was designed for
NT4.0, and tested under NT4.0 SP3-SP6, 2000 and XP. The side effect 
of installing custom GINA under Windows XP is changing of the login desktop 
appearance. In any case, you had better to run RDISK.EXE
to update repair info and make repair disk BEFORE the installation. 

3.11 The program uses USERPROFILE environment variable to get path
to user's directory for saving data.