User Manual

92
more words to be accepted.
For example, in applications
where the operator is
constantly talking, changing
this parameter to a negative
number (-0.5 is a good number
to start with) may result in
better overall system
recognition performance. The
tradeoff is that out of
vocabulary words and
extraneous noises are more
likely to be erroneously
recognized as words.
BlueStreak_Frontend_
PrintBackgroundInterval
How often, in seconds, the
background energy is printed
out in debug.
10
BlueStreak_Train_
EmbeddedMiddleWord
When doing catchweights (or
some other rapid-digit task)
and there is a word that is
often mixed in with the digits
(such as point or decimal),
then that word should be
specified in this parameter.
The word will then be trained
in an embedded context with
the digits, improving
recognition performance on
catchweights. To use no
embedded word, set this
parameter to NONE.
. (i.e. the default is "point")
BlueStreak_Train_
MaximumUttSeconds
The maximum length, in
seconds, of a training
utterance. Utterances longer
than this value will be
discarded.
The value for this parameter
must be increased when
extremely long vocabulary
words are used in a task. This
value can also be decreased to
help ensure proper training
data when only shorter words
are used.
Minimum value = 1
Maximum value = 10
7
BlueStreak_Train_
RandomizeEnable
Gives the system administrator
the ability to randomize
training the same way or
differently every time an
operator goes through the
enrollment training process.
0 = randomize training the
same way every time
1 = randomize training
differently every time