Skip to main content Link Search Menu Expand Document (external link)
Hidale.com Docs

@speech attribute

🔌dp.kinect dp.kinect2
signature
speech INT
values
0
off default
1
recognition
2
recognition and interference
examples
@speech 1

Speech recognition and speech interference output on the fifth outlet. It is optimized for command-response usage. It does not support dictation. The sensor’s microphone array allows localizing the speaker and hearing them without wearing a headset.

The speech recognition grammar (the words to recognize) are defined in a GRXML file as defined by the W3C standards. The speech recognition does support multiple simultaneous languages. US English is installed with the official drivers. Additional language packs are available.

# @skeletonformat 0
/speech/recognize/ruleid tag conf srconf heardwords
/speech/error/interfere problem

# @skeletonformat 1
speech recognize ruleid tag conf srconf heardwords 
speech error interfere problem
  • ruleid is the value of the id attribute on the <rule> tag within the GRXML file
  • tag is the text between the <tag> tags within the GRXML file
  • conf is a floating point number [0.0, 0.5, 1.0] representing the increasing confidence in recognition. This confidence is used for filtering via the @speechconf attribute.
  • srconf is a floating point number inclusively [0.0 - 1.0] and specific to the sensor recognition technology. It can be used for more detailed confidence information.
  • heardwords is a single symbol containing the entire phrase recognized and matched to a tag. If you defined an <item> in your GRXML file which contained several words in a single phrase to be recognized, then heardwords will be a single symbol containing all of those defined words separated by a space character.
  • problem is a single symbol representing the interference to speech recognition. These error messages are optional and only enabled when @speech 2.

The possible symbol values for interference problems are:

  • noise The sound received is interpreted by the speech recognition engine as noise.
  • nosignal A sound is received but it is of a constant intensity. This also includes the microphone being unplugged or muted.
  • tooloud A sound is received but the intensity is too high for recognition. If you receive many of these events, it could be due to a known issue with the sensors’s default audio gain settings. Try enabling the @autogain attribute. Or, for more explicit fine-tuned control, open the Windows Control Panel and select Sound, then the Recording tab. Select your sensor (e.g. Kinect Microphone Array) and then press the Properties button. Select the Levels tab. Set the Microphone Array gain level to a lower value. Press OK and then OK.
  • tooquiet A sound is received but the intensity is too low for recognition.
  • toofast The words are spoken too quickly for recognition.
  • tooslow The words are spoken too slowly and indicates excessive time between words.