annotate_INTSINT_MOMEL: Automatic annotation of intonation using MOMEL and INTSINT...

In humlab-speech/reindeer: This package extends the capabilities of emuR to work with speech data in a nordic climate

Automatic annotation of intonation using MOMEL and INTSINT labels

Description

This function transcribes the utterances specified in a segment list using the MOMEL/INTSINT framework \insertCiteHirst:2005jg,Hirst:2011wqreindeer and insert the EVENT labels into a special tier ("Intsint" by default). The function will then call the Praat routines \insertCiteHirst:2007tyreindeer and the external 'momel' program to make the calculations, and insert the resulting levels onto the transcription tier. Please note that calling multiple external programs results in some overhead, so please allow some time to complete the transcription for large sets of utterances.

Usage

annotate_INTSINT_MOMEL(
  emuDBhandle,
  seglist,
  windowSize = 30,
  minF = 60,
  maxF = 750,
  pitchSpan = 1.5,
  return.wide = TRUE,
  maximumError = 1.04,
  reducWindowSize = 20,
  minimalDistance = 20,
  minimalFrequencyRatio = 0.05,
  intsint.level = "Intsint",
  attribute.definitions = c("Momel", "Intsint", "IntsintMomel"),
  absolute.tone.labels = c("T", "M", "B"),
  relative.tone.labels = c("H", "U", "S", "D", "L"),
  force = FALSE,
  verbose = FALSE,
  praat_path = NULL
)

Arguments

emuDBhandle	The Emu database handle.
seglist	A segment list resulting from a query function call, containing utterances that should be transcribed.
windowSize	The window size (in ms) used when extracting f0. One hald window size of duration will also be added to each end of all extracted utterance segments to allow f0 target points to be discovered right at the beginning and end of utterances.
minF	The low end of the fundamental frequency range to consider when searching for a pitch contour.
maxF	The high end of the fundamental frequency range to consider when searching for a pitch contour.
pitchSpan	The maximum pitch span to consider. Defaults to 1.5 octave, but could be expanded to 2.5 for expressive speech.
return.wide	Should the output be formated as a long tibble, with all attribute names in a column 'attribute' and labels in a column 'labels', or in wide format (where all attributes are given their own column).
maximumError	The maximum error allowed by MOMEL. See the momel source code for more information.
reducWindowSize	The size of the reduced analysis window used by MOMEL. See the momel source code for more information.
minimalDistance	The minimal distance between INTSINT labels. See the momel source code for more information.
minimalFrequencyRatio	See the momel source code for information on this parameter.
intsint.level	THe name of the transcription level that should be created and used for storing annotations. If this level already exists, it will be used if attribute definitions match the ones given here.
attribute.definitions	The names of attribute definitions to set up for the transcription level. Currently not used.
absolute.tone.labels	The INTSINT labels that should be set up as members of the 'Absolute_tones' label group.
relative.tone.labels	The INTSINT labels that should be set up as members of the 'Relative_tones' label group.
force	Force insertion of transcriptons. Currently not implemented.
verbose	Should output be displayed for all major steps?
praat_path	An explicit path to the Praat executable. Usually not required.

Details

As a side effect, this function also sets up two label groups on the attribute definition in which INTSINT labels are inserted (default "Intsint"):

1. Absolute_tones (T,M,B), and 2. Relative_tones (H,U,S,D,L)

which may be used in queries. In order to use them in larger queries, the user has to define links between the created level and existing transcription levels using add_linkDefinition, and links between created labels on the two tiers using autobuild_linkFromTimes.

At the start of the MOMEL procedure, Pitch will computed using a wide frequency search. The user may guide the search by supplying more restrictive maximum and minimum fundamental frequency values (minF, maxF). By default, the search considers a moderate pitch span (pitchSpan=1.5), but if very expressive speech is investigated, pitchSpan may be increased to 2.5.

The MOMEL / INTSINT procedure also returns the MOMEL deduced f0 target associated with the INTSINT label, as well as the f0 predicted by the INTSINT model. These are returned by the function but currently not inserted as labels in the transcriptions.

Value

A tibble containing computed INTSINT (and MOMEL) information for each bundle, with start time relative to the start of the bundle media file.

References

\insertAllCited

Examples

## Not run: 
reindeer:::unlink_emuRDemoDir()
reindeer:::create_ae_db(verbose = TRUE) -> emuDBhandle
add_ssffTrackDefinition(emuDBhandle,"f0",onTheFlyFunctionName = "ksvF0")
query(emuDBhandle,"Intonational =~ .*") -> sl
annotate_INTSINT_MOMEL(emuDBhandle,sl) -> momelTab
add_perspective(emuDBhandle, "INTSINT")
set_signalCanvasesOrder(emuDBhandle,"INTSINT",c("OSCI","SPEC","f0"))
set_levelCanvasesOrder(emuDBhandle,"INTSINT",c("Intsint"))
serve(emuDBhandle)

## End(Not run)

humlab-speech/reindeer documentation built on May 21, 2023, 4:43 p.m.