Short Description of the Automatic Prosodic Aligner (ProsAlign)
This is a short description of the automatic prosodic aligner (ProsAlign) program. This
program may be used to label automatically prosodic events in a speech file.
From the speech file the parameters F0, voicing and RMS-amplitude are extraced
and used as input for the program. The input is therefore a structured file
with continuous values for F0, voicing, and RMS-amplitude (usually an
ESPS/waves *.f0-file). The output will be a structured label file with the
type of prosodic event and its position in time.
The program is currently in development but already usable as the illustration
of an output created by the prosalign program shows below. Here the
automatically generated label file can be compared with the manually established
one (example taken from the GToBI
training material). The following three
label files are depicted:
1. Words
2. automatically generated pitch accents and boundary tones
3. manually generated pitch accents and boundary tones.
The taskflow within the program is shown in the following picture:
