SSSHP Contents | Labs

 MATTINGLY 1974, p. 2475 
Go to Page | Contents Prosodic Features | Bibl. | Page- | Page+
 

of a phone to determine duration, fundamental frequency and intensity.

In our own prosodic control scheme for British English (Mattingly 1966), two degrees of stress and three common intonation contours (fall, fall-rise and rise) can be marked in the input. The Fo rules specify a falling contour during the 'head' of the breath group; the slope varies with the quality of the syllable nucleus. Voiceless consonants cause a 'pitch skip'; stressed syllables, a smooth rise in Fo. The required terminal intonation contour is imposed on the 'tail' -- the last stressed syllable and any following syllables. Each possible syllable nucleus has an inherent duration which is increased multiplicatively by stress. In prepausal syllables, the duration of all phones is increased and amplitude is gradually diminished. More recently (Mattingly 1968a), we have used similar rules for synthesis of General American, and in addition, provided for the durational effects of juncture.

Rabiner (1969) follows the model proposed by Lieberman (1967) in which the overall fundamental contour is determined by subglottal air pressure, except for the so-called 'marked' breath group where laryngeal tensing produces a terminal rising contour. Four degrees of stress can be indicated; the higher the stress, the greater the increase in Fo on the stressed syllable. Duration increases additively with the openness and tenseness of the vowel and the degree of stress, as well as being affected by the following consonant.

Hiki and Oizumi (1967) have developed prosodic rules similar to those of Rabiner and Mattingly. The Fo rules deal with pitch accent, emphasis, terminal contours, the overall contour and individual differences; the duration rules take into account the inherent duration of phones, pause length and accent; and interestingly, the effect of changes of tempo on these features.

Umeda et al. (1968) determine prosodic patterns for English directly from ordinary printed text and use them to control the vocal-tract synthesis by rule scheme of Matsui (1968). Syntactic rules of a primitive kind are used to divide an utterance into blocks corresponding to breath groups. An overall Fo contour is imposed on each breath group. Word stress (along with the phonetic transcription for a word) is determined by table lookup, and the fundamental frequency and duration are increased accordingly. Intonation contours and pause duration are derived from the punctuation, if any, following each block.

Vanderslice (1968) has also considered prosodic feature from the standpoint of the problems involved in the conversion of orthographic text to sound, correctly distinguishing some features not included in earlier systems and proposing rules for their synthesis. Two degrees of pitch prominence are used instead of one: 'accent' and 'emphasis', the latter for contrastive and emphatic stress. The features 'cadence' and 'endglide' replace the traditional fall, fall-rise and rise, cadence being equivalent to a fall, end-glide to a rise, and cadence followed by end-glide to fall-rise. 'Pause' is a separate feature. To account for the raising of Fo in quoted material and its lowering in parenthetical material, the features 'upshift' and 'downshift' respectively are used. An additional group of 'indexical' features is proposed
 

Go to Page | Contents Prosodic Features | Bibl. | Page- | Page+

 MATTINGLY 1974, p. 2475 
SSSHP Contents | Labs
Smithsonian Speech Synthesis History Project
National Museum of American History | Archives Center
Smithsonian Institution | Privacy | Terms of Use