4th International Conference on Spoken Language Processing

Philadelphia, PA, USA
October 3-6, 1996

On the Quantal Nature of Speech Timing

Gunnar Fant, Anita Kruckenberg

Dept. of Speech, Music and Hearing, KTH, Stockholm, Sweden

This is a review of regularities we have observed in the analysis of text reading, mostly Swedish, directed to the timing of vowels and consonants, syllables, interstress intervals and pauses. We have found tendencies of quantal aspects of temporal structure, superimposed on more gradual variations, which add quasi-rhythmical elements to speech. A local average of interstress intervals of the order of 0.5 sec appears to function as a reference quantum for the planning of pause durations. A recent study, confirming our previous findings of multiple peaks with about 0.5 sec spacing in histograms of pause durations, provides support to this model. It is well established that pause durations tend to increase with increasing syntactic level of boundaries. However, these variations tend to be quantally scaled even within a specific boundary category, e.g. between sentences or between paragraphs. Relatively short pauses, as between phrases or clauses, show durations in complementary relation to terminal lengthening. There are indications of approximately 1, 1/2, 1/4, 1/8 ratios of average durations of interstress intervals, stressed syllables, unstressed syllables and phoneme segments which adds to the observed regularities. The timing of syllables and phonetic segments with due regard to relative distinctiveness and reading speed will be discussed and also tempovariations within a sentence.

Full Paper

Bibliographic reference.  Fant, Gunnar / Kruckenberg, Anita (1996): "On the quantal nature of speech timing", In ICSLP-1996, 2044-2047.