JETI model DUPLEX DC-24, Руководство пользователя, Страница: 117 / 168

computer radio control system EN
Voice recognition modes
Push-To-Talk mode – a)
The voice command must be
pronounced while the Push- To-Talk
switch is being held. This method is
more immune to recognition errors
since it only expects speech after
the Push-To-Talk switch is pressed.
b) Keyword mode (always
listening) – The voice command
recognition is started after the
k e y p h r a s e i s s u c c e s s f u l l y
recognized. That means first, you
say the keyphrase and the
transmitter beeps for confirmation
that it understood. Then you say the
required command phrase. Since
the transmitter uses voice activity detection, you don’t need to push
any buttons. You must say the command within 2 seconds after the
transmitter beeps, otherwise it won’t be accepted.
Teaching the phrases
The transmitter must learn the voice phrases before they can be
recognized. Once the phrase has been successfully trained it can be
used among all models since the voice commands are shared
between each other. The most important is to provide the
transmitter the best audio samples. Please be sure that you are
located in a quiet room without sound disturbances that could
influence the learning process.
Locate the Voice Commands menu, choose the Keyword mode. 1.
Open the Keyword settings dialogue. Here assign the Push-To- 2.
Talk switch. Choosing the switch is important even in the Keyword
mode since the switch will mark the beginning and the end of
speech phrase. We recommend using a spring loaded switch for this
purpose.
Push the switch and pronounce the command. After the Push-
3.
To-Talk switch is activated, the transmitter waits for the incoming
speech. As soon as the voice activity is detected, a small rectangle
will be shown at the top bar of the display. Pronounce the phrase like
you’d do it at the airfield but loudly and clearly.
Release the switch after you’ve finished speaking. The
4.
command template will be stored to the SD card and a double-beep
will confirm that the process has been successful.
You need to perform the training phase 3-times to get all 3
5.
templates with the same phrase. This procedure will improve the
recognition accuracy.
Recognizing the phrases
You can verify that the correct phrases have as low scores as possible
by pressing the
"F1” Dots button in the Voice Commands menu.
What you see is a scoring table where all samples are compared
against the current speech sample. A low scores mean that the
samples are similar to each other. On the other hand, high scores are
results of dissimilar samples. For each command, all three scores
should not exceed certain sensitivity that can be set in the voice
command detail.Otherwise the command will not be recognized.
4. 4.
115