TTS-driven Synthetic Behavior Generation Model for Embodied Conversational Agents - Coverbal Synchrony in Human-Machine Interaction - page 344

2, are determined by the phase tagging deque as: ' vedno iste' ( always
the same ) (CU-1), and ' samo iste obraze ' ( only the same faces ) (CU-2).
The prosodic word phrases, as determined by the predicted phrase
breaks, demand some adaptation of these content units. For example,
the predicted word phrases are ' vedno iste ' (WP-1), and ' in samo iste
obraze' (WP-2). The fi rst Content unit CU-1 obviously matches the WP-
1, whereas the second Content unit CU-2 has to be extended to also
include the word ' in '. When, however, the CUs include more words
than the predicted WPs, those words are automatically disregarded. The
second process ' Searching of emphasized words' identifi es the emphasized
words. As the emphasized word is identifi ed, those words contain a
syllable assigned with a PA label (primary accent). The primary accent
is assigned by the symbolic prosody deque (using CART models) to
the most accentuated syllable inside the intonation prosodic phrase.
In the context of movement phases, the stroke-movement phase is
performed, co-aligned with the word containing the PA syllable. When,
however, the CU has no PA syllable (and no emphasized word), the
CU is disregarded. Further, the words preceding the emphasized word
represent preparation, or pre-stroke-hold movement-phases. And words
following the emphasized word represent the post-stroke-hold, or
retraction movement-phases. On other syllables within the intonation
prosodic phrase, the secondary accents (NA) are predicted. The NA
prominence labels in general identify the preparation movement-phase
(when the word that contains the NA syllable precedes the prosodic
gesture trigger), or the retraction movement-phase (when this is the
last coverbal gesture in the sequence, or a longer pause is indicated).
Finally, the silences ( sil ), when predicted by the acoustic prosody deque ,
indicate the existence of the hold-movement-phases. The third process
' Searching of stressed syllables ' aligns the stroke movement phases with
speech fragments, according to the PA syllable. The starting and ending
points of the stroke phase are determined by the beginning of the
emphasized word, and by the end of the PA syllable. In Figure 6, there
are two emphasized words, each containing a PA syllable. The fi rst PA
syllable ' is ' (emphasized word ' iste ' ( the same )) identifi es the position
of the fi rst meaningful shape. The second PA syllable ' mo ' (emphasized
word ' samo ') identifi es the position regarding the manifestation of the
second meaningful shape. The fourth process ' Align stroke ' then aligns
the stroke movement-phases to the speech fragments, as indicated
by the stressed syllables. The fi fth process ' Align preparation ' aligns
the preparation movement phase. The starting and ending points of
the preparation phase are identifi ed by the NA syllable, and the end
of the 'preparation' word. In Figure 6, the words ' vedno ' and ' in ' are
indicated as 'preparation' words for the fi rst and second prosodic word