Detail of contribution
Auteur: Anatole FIODOROV
Titre:
Idiosyncrasy of Speech: Professor D. Crystal's Prosodic Profile
Abstract/Résumé: The problem of speaker’s idiosyncrasy and forensic studies are closely related. The researchers will never cease to look for parameters that may help to reliably identify the speaker both statistically and acoustically [1, 2]. In the present study we used speech material produced by Prof. Crystal (DC), and speaker GD, a linguist, both native speakers of English. They delivered three lecturettes for the BBC. The speakers were the authors of their texts which were read in an informal manner. The recordings were segmented manually into words and measured using audio feedback and spectrograms with the help of PRAAT program [3]. The duration of segments was calculated in ms (±5-10 ms) without pauses and false starts. On the basis of standard phonemic transcription the average duration of a syllable was worked out as the harmonic mean. Syllable absolute durations were then determined in relation to the total harmonic mean for each speaker. The dynamics of syllabic relative timing were shown on charts which allowed to ascertain syntagms and measure their duration. A few reiterated words and syntagms were analysed in terms of formant and prosody movement. The descriptive statistics like the mean duration of syllable (203 ms), standard error of the mean (7.6ms) and variation coefficient (53.3%) have proved to be significantly indicative for speaker DC. The spectrograms of reiterated words ‘language’ and ‘spam’ have shown a rather marked variation, though ‘language’ being much more indicative. Some insight into the reiterated F0 configurations of certain syntactic structures has been gained. These data contrasted significantly with the other subject’s statistics. The syntactic structure of the text along with the formant tracking, F0 analysis and timing of the prepared speech contribute significantly to the idiosyncrasy of the speaker. The data have lead to believe that speaker DC controls negligible changes in relative syllabic durations to lend an emphasis to a word in the syntagm. The study suggests that certain F0 syntactic configurations are fairly informative, however the timing parameters prove to be more idiosyncratic and syntactically bound. References [1] Nolan, F. (1997) Speaker Recognition and Forensic Phonetics, The Handbook of Phonetic Sciences, (eds. W.Hardcastle and J.Laver), Blackwell Publishers Ltd, 744-767. [2] Zellner Keller, B. (2002) Revisiting the Status of Speech Rhythm, Proc. of Speech Prosody Conference 2002 (eds. B. Bel and I. Marlien), Aix-en-Provence, April 11-13, 2002, 727-730. [3] Boersma, P., Weenink, D. (2012) Praat: doing phonetics by computer. Version 5.3.17, retrieved 6 January 2012 from http://www.praat.org/
Titre:
Idiosyncrasy of Speech: Professor D. Crystal's Prosodic Profile
Abstract/Résumé: The problem of speaker’s idiosyncrasy and forensic studies are closely related. The researchers will never cease to look for parameters that may help to reliably identify the speaker both statistically and acoustically [1, 2]. In the present study we used speech material produced by Prof. Crystal (DC), and speaker GD, a linguist, both native speakers of English. They delivered three lecturettes for the BBC. The speakers were the authors of their texts which were read in an informal manner. The recordings were segmented manually into words and measured using audio feedback and spectrograms with the help of PRAAT program [3]. The duration of segments was calculated in ms (±5-10 ms) without pauses and false starts. On the basis of standard phonemic transcription the average duration of a syllable was worked out as the harmonic mean. Syllable absolute durations were then determined in relation to the total harmonic mean for each speaker. The dynamics of syllabic relative timing were shown on charts which allowed to ascertain syntagms and measure their duration. A few reiterated words and syntagms were analysed in terms of formant and prosody movement. The descriptive statistics like the mean duration of syllable (203 ms), standard error of the mean (7.6ms) and variation coefficient (53.3%) have proved to be significantly indicative for speaker DC. The spectrograms of reiterated words ‘language’ and ‘spam’ have shown a rather marked variation, though ‘language’ being much more indicative. Some insight into the reiterated F0 configurations of certain syntactic structures has been gained. These data contrasted significantly with the other subject’s statistics. The syntactic structure of the text along with the formant tracking, F0 analysis and timing of the prepared speech contribute significantly to the idiosyncrasy of the speaker. The data have lead to believe that speaker DC controls negligible changes in relative syllabic durations to lend an emphasis to a word in the syntagm. The study suggests that certain F0 syntactic configurations are fairly informative, however the timing parameters prove to be more idiosyncratic and syntactically bound. References [1] Nolan, F. (1997) Speaker Recognition and Forensic Phonetics, The Handbook of Phonetic Sciences, (eds. W.Hardcastle and J.Laver), Blackwell Publishers Ltd, 744-767. [2] Zellner Keller, B. (2002) Revisiting the Status of Speech Rhythm, Proc. of Speech Prosody Conference 2002 (eds. B. Bel and I. Marlien), Aix-en-Provence, April 11-13, 2002, 727-730. [3] Boersma, P., Weenink, D. (2012) Praat: doing phonetics by computer. Version 5.3.17, retrieved 6 January 2012 from http://www.praat.org/