Phones & Phonemes 2025-04-09 |
Dieter Maurer, Prof. emerit. Dr. University of Zurich, Department of Computational Linguistics Guest, Phonetic and Speech Sciences |
||
Home
Research summary News/Check online Publications Contact |
Research summary In our research on vowel acoustics, when re-examining basic experiments reported in the literature but also including an extensive variation of sound production parameters, several findings proved to be opposed to the thesis of vowel-specific patterns of formants or resonances (the main thesis of the prevailing theory). We were able to demonstrate that: (i) The vowel spectrum relates to fundamental frequency (fo). (ii) Formant patterns or spectral shapes are ambiguous, representing two or three vowel qualities if fo is varied. (iii) Speech, minimal pairs and isolated vowel sounds produced at fo equalling or exceeding the first formant frequencies as given in formant statistics are recognisable. (iv) The spectra of sounds of a vowel vary in a nonuniform way. (v) Synthesised vowel sounds with spectral characteristics outside the framework of formants or resonances are recognisable. (vi) The vowel quality is not lost but shifts for low-pass and high-pass filtered sounds, (vii) Actual vocal tract configurations are not always directly related to measured formant patterns. Most of these findings and the related references are described and documented in a first book on vowel acoustics [1] as a compilation of the results of single published papers (see the Publications menu item). Based on these findings, we continued to address two main research topics: The creation of a large sound database that can serve as a reference for verification or falsification of any thesis on vowel acoustics, and the formulation of knowledge-based general statements that apply to the vowel spectrum of all recognisable vowel sounds independent of their production parameters. Hence, we have created the Zurich Corpus of Vowel and Voice Quality [2] which, in its second version [3], comprises more than 37,700 recordings. Subsequently, in an extensive investigation of this corpus, we were able to formulate three general statements concerning vowel acoustics: The vowel sound and its recognised vowel quality do not relate to measured fo but to pitch. The vowel sound is a perceptual and acoustic foreground–background. phenomenon. The spectral representation of vowel quality is nonuniform. Above all, the evidence provided for a vowel–pitch relation represents an unexpected scientific discovery and calls for future research on the role of perception in vocal production and a corresponding theoretical approach to vowel acoustics. Extensive documentation and discussion of the experiments conducted, and results obtained, including links to the investigated sounds documented in the Zurich Corpus are published in a new book on vowel acoustics in 2024 [4]. References [1] Maurer, D. (2016). Acoustics of the Vowel – Preliminaries. Peter Lang CH. >> DOI: 10.3726/978-3-0343-2391-8 (download link) [2] Maurer, D., d’Heureuse, C., Suter, H., Dellwo, V., Friedrichs, D., & Kathiresan, T. (2018, September). The Zurich Corpus of Vowel and Voice Quality, Version 1.0 [sound archive]. Interspeech 2018, 1417–1421. Retrieved June 14, 2018, from https://zhcorpus.org DOI 10.21437/Interspeech.2018-1542 >> DOI: 10.21437/Interspeech.2018-1542 (download link) [3] Maurer, D., d’Heureuse, C., Suter, H., Dellwo, V., Friedrichs, D., & Kathiresan, T. (2024, July 31). The Zurich Corpus of Vowel and Voice Quality, Version 2.0 [sound archive]. Retrieved April 9, 2025, from https://zhcorpus.org >> The Zurich Corpus of Vowel and Voice Quality [4] Maurer, D. (2024). Acoustics of the Vowel – Indices. Peter Lang CH. DOI 10.3726/b22003 >> DOI: 10.3726/b22003 (download link) |