^SynOhrMultiSenseKNX-Eng-13.odt, 2014-05-06 Page 36 of 49
Configuration
of the Speech
Recognizer
Range
The maximum range of the speech recognition is 7 m at normal speaker volume. The range is
specified in SynOhrStudio via parameter slider „Sensitivity“ (see Picture 22). The larger the room
size is set the more sensitive are the microphones. This is noticeable in accordance with the
sound level monitoring. Increased range means under circumstances decreased quality of the
speech recognition, e.g. if a playing radio is close to the SynOhr MultiSense KNX but the
speaker a few meters away.
Master Slave
For larger living rooms we therefore recommend a master slave operation with multiple stations,
each set only to a standard range of about 4 m. Via the KNX
TM
Bus and an additonal required
Enertex® EibPC the signals can be evaluated and the speech recognitions can be synchronized
with each other.
Keyword
Operation
In „SynOhrStudio“ the independent application from ETS the SprachSchatz for the commands is
put together respectively compiled. One command consists of maximum four words. Under one
command is understandable e.g. WALL LAMP BRIGHTER. So that SynOhr MultiSense KNX
does not recognize a command during a conversation and falsely enables an action, the
SynOhr®SynOhr MultiSense KNX must be placed in the active mode first. This is also done with
the help of speech recognition using the „Keyword“. This keyword is intensively checked by the
recognizer and must be spoken isolated (pause at the end). If the recognizer has realized the
keyword, so it gives a visual feedback through yellow flashing the ring and audible feedback by
playing a short ring tone. During the phase of blinking and after playing the feedback sound the
actual command must be spoken within three seconds.
Selection of
Keywords
The keywords are selected so that these have different properties for recognition and stability.
The artificial word ENERTEX or the word COMMAND are from our experience as particularly
suitable. Should children operate the SynOhr MultiSense KNX by speech recognition before
voice break, so is probably SYNOHR the best choice.
Sensitivity
With this setting the maximum range for speech recognition is set. The maximum range depends
mainly on the room acoustics, the place of installation (e.g. acoustic shadow behind a wall), the
speaker volume and (but less) of the number of used words. Up to a range of 4 m should a
suitable everyday use parameterization even under difficult circumstances. For the range it is
important whether it is spoken frontally or from one side. Spoken laterally the signal is more than
50 % degraded. Therefore the human being automatically turns the head towards the speaker.
Additionally the similarity of the selected words and the room acoustic plays a large role in
theses distances. If the noise sources (radio etc.) are louder and closer to the SynOhr than the
speaker a recognition is not possible. If two people have a conversation a third person cannot
operate the SynOhr from the same distance. Also note the chapter in the manual for choosing a
suitable location. In order to „listen“ acoustically larger areas, it definitely takes more than one
slave.
The following table specifies the operating range adjustment. The situation with a lot of
resonances are often to be found in the corridor or rooms with lots of tiles and no wallpaper.
Incorrect parameter here means that very probably a recognition about possibly greater
distances are possible, but this is not robust against interference noise.
Because in hallways or tiled rooms and without wallpapers echo and resonances are existing,
you have to work with smaller „Room Size“ as in rooms with sound-absorbing elements (wooden
Enertex
®
Bayern GmbH – Ebermannstädter Straße 8 - 91301 Forchheim - [email protected]
