You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
43 lines
2.2 KiB
43 lines
2.2 KiB
9
|
|
|
|
The prosodic component
|
|
|
|
9.1 Overview
|
|
|
|
The sentence representation produced by the phonological component PHONO2
|
|
serves as input to the prosodic component PROSOD that is to be described in this
|
|
chapter. An example of the input to the prosodic component and the output
|
|
generated by the prosodic rules is shown in Figure 9-1. The output consists of a
|
|
string of phonetic segments, with each segment assigned a stress feature and a
|
|
duration in msec. The fundamental frequency targets which appear in the
|
|
PROSOD output listing are generated by an obsolete algorithm and are discarded
|
|
by FOTARG which then generates the proper FO targets.
|
|
|
|
9.2 Segmental durations
|
|
|
|
In a review of the factors that influence segmental durations in spoken English sen-
|
|
tences (Klatt, 1976b and references cited therein), it was concluded that only a few
|
|
of the many rule-governed durational changes are large enough to be perceptually
|
|
discriminable. The goal of the rule system described below and in Klatt (1979b) is
|
|
to characterize these perceptually important first-order effects.
|
|
|
|
The durational definitions that have been adopted include the closure for a
|
|
stop (any burst and aspiration at release are assumed to be a part of the following
|
|
segment). For fricatives, the duration corresponds to the interval of visible frica-
|
|
tion noise (or to changes in the voicing source if no frication is visible). For
|
|
sonorant sequences, the segmental boundary is defined to be the half-way point in
|
|
the formant transition for that formant having the greatest extent of transition.
|
|
These definitions lead to a convenient and largely reproducible measurement pro-
|
|
cedure, but the physiological and perceptual validity of these boundaries have not
|
|
been established.
|
|
|
|
Each segment is assigned a duration by a set of rules presented in detail
|
|
below. The rules are intended to match observed durations for a single speaker
|
|
(DHK) reading paragraph-length materials. The rules operate within the
|
|
framework of a model of durational behavior which states that: 1) each rule tries to
|
|
effect a percentage increase or decrease in the duration of the segment, but 2) seg-
|
|
ments cannot be compressed shorter than a certain minimum duration (Klatt,
|
|
1973). The model is summarized by the formula:
|
|
|
|
93
|