^SynOhrMultiSenseKNX-Eng-13.odt, 2014-05-06 Page 46 of 49
In summary:
1. The keyword must be parameterized robust, so that
2. the command recognition guarantees with the help of the similarity strategy a high rate
of recognition.
3. The room size may not be chosen too large in reverberant ambient, otherwise the
keyword recognition is wrong.
Both the similarity strategy including the robustness can be customized by the user to his own
needs.
Experts
Settings
Basics
In addition the menu
Options
an expert mode (
Experts)
is adjustable. This opens up additional
settings. With an unfavorable choice of settings the recognition of words can be impossible.
Therefore the change of parameters is recommend only to experts. In addition the parameter
may be modified only in small steps and their effect can be directly tested in the object.
Noise Level
The noise level of background noise in the room can be adjusted via the menu item
Noise Level
From silent (Silent) up to very loud (Extreme) are nine levels adjustable. This setting chances
the dynamics of the recognition. Basically the speaker must stand out in the higher level
significant from his surrounding, so that a good recognition is guaranteed. Conversely it may be
easier for the user in a soundproofed rooms and very quiet rooms (Silent) the speech
recognition to operate at an appropriate parameterization.
In rare cases and depending on the room acoustics there will be false alarms in children's- or
women's voices. In this case set
Noise Level
on Talking or – if necessary – Loud.
Feedback Time
The length of time until the confirmation to the command is set via the menu item „Feedback
Time“. Overall five levels are adjustable from time-lag (Paralyzed) up to fast (Fast). The
feedback time is the time of break, which the user must wait after spoken the keyword until the
speech recognizer switches in the active mode. A longer feedback time means a higher
robustness for the recognizer against disturbing ambient noises.
Ready Signal
In this menu item can be set whether an audible signal upon activation of the SynOhr
MultiSense KNX (Ready Signal) is to be output. The default value is set to 'Yes'. The sound can
be adjusted by individual modifying the file
t_ready.wav
on the SD card.
Feedback Signal
In this menu item can be set whether an audible signal to confirm a command sequence
(feedback signal) is to be output. The default value is set to 'Yes'. The sound can be adjusted by
individual modifying the file
t_exec.wav
on the SD card.
Dynamics
Speakers tend to have a certain dynamics, that means they get lower after the keyword. In order
to counteract this, SynOhr MultiSense KNX can automatically increase the sensitivity of the
microphone after the spoken keyword from Off (no increase) up to maximum (maximum
increase). This improves the recognition of commands in a rather quiet ambient.
Speech Quality
Activation
In the menu item „Speech Quality Activation“ can be adjusted, how precisely the keyword for the
activation of the speech switch must be spoken and indeed from the value „Everything“ up to
„Exact“ with increasing accuracy. Similar to the parameter „Clarity“ (comp. page 37) causes this
in the speech recognizer a increased accuracy requirement when speaking the keyword.
Here is used a deviating algorithm, which has an influence mainly on the recognition of
Enertex
®
Bayern GmbH – Ebermannstädter Straße 8 - 91301 Forchheim - [email protected]
