User Manual

122
the terminal then proceeds with the task from the point at which you left off.
Note: The length of time that it takes for the terminal to compute the model for the word or digit that
has been retrained is approximately five seconds for a word and twenty seconds for a digit.
Note: When an operator is retraining a digit, the operator will not only be prompted
to repeat the digit four times (as opposed to ten times for regular words), but he/she
will also be prompted to say some of the embedded digit strings that contain the
particular digit. Embedded digit strings are strings of numbers that contain the digit
that is being trained. For example, for the digit "2", the embedded strings might be
020, 212, and 222. When training the embedded digit strings , the digits should be
spoken without pauses between them. The operator should speak the digits as
naturally as he/she would when telling someone a phone number.
Also, if the task that is being run contains decimal points (i.e. the word "point")
and the operator chooses to retrain a digit, the embedded digit strings will contain a
string with the word "point" in it (i.e. 2 point 2).
Sampling Noise with Talkman® T2
Note: The information in this topic applies specifically to the Talkman® T2 system.
To view this information for the Talkman® OPEN system, click on the image of the
terminal.
• Purpose
• Procedure
Purpose
This document contains instructions for taking a noise sample for a Talkman® T2 terminal. You
may need to take a noise sample if the terminal is having trouble hearing you because of
changes in the noise level in your working environment.
If the terminal is having trouble with only a word or two and taking a background noise sample
does not improve the terminal's recognition, you may need to retrain the word or words that
the terminal is having trouble recognizing.
Notes:
•
A noise sample is taken by having the operator repeat the digits '0' through '4'. If
the terminal does not hear the operator say one of the digits, it will say, "I didn't
hear you, please speak up a bit." The terminal will then have the operator repeat
the digit that it couldn't hear before continuing with the rest of the noise sample .
•
If the terminal completes the noise sample and determines that the input volume is
unacceptable, it may ask the operator to move the microphone away from his or
her mouth by saying, "Please move microphone farther away from your mouth."
•
During a noise sample , if the terminal detects a sound that is loud enough to be
considered speech and that lasts longer than two seconds, the terminal says,
"Ignoring last phrase. Too long." When the terminal can no longer detect the sound,
it will ask the operator to repeat the digit that just failed. This feature has been
designed to handle scenarios that could adversely affect a noise sample , such as
having the PA system go off while performing a noise sample . However, since the
terminal must wait until the sound has ended before continuing the noise sample