Cepstral coefficients successfully distinguish the front Greek fricatives
In the current study, we explore the factors underlying the well-known difficulty in acoustic classification of front fricatives (McMurray & Jongman, 2011; Maniwa et al., 2009) by taking a closer look at the production of 29 native Greek speakers. Our corpus consists of Greek fricatives from fiv...
Gespeichert in:
Veröffentlicht in: | The Journal of the Acoustical Society of America 2018-09, Vol.144 (3), p.1906-1906 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | eng |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | In the current study, we explore the factors underlying the well-known difficulty in acoustic classification of front fricatives (McMurray & Jongman, 2011; Maniwa et al., 2009) by taking a closer look at the production of 29 native Greek speakers. Our corpus consists of Greek fricatives from five places of articulation and two voicing values [f, v, θ, ð, s, z, ç, ʝ, x, ɣ] produced in nonce disyllabic words before [a, o, u] in stressed syllables (e.g., [ˈθakos]). We apply a relatively novel classification method based on cepstral coefficients, previously successful with obstruent bursts (Bunnell et al., 2004), vowels (Ferragne & Pellegrino, 2010), and Romanian fricatives (Spinu & Lilley, 2016). Our method yields the best correct classification rates reported to date for front fricatives: Present study: 88%; English: 66% (Jongman et al., 2000), Greek: 55.1% (Nirgianaki, 2014). The important cues for the successful classification are the vowel following the target fricative and the second region of frication noise. Our study adds to the body of work aimed at identifying techniques for quantifying and categorizing large samples of speech. Obtaining higher classification rates than before takes us one step closer to understanding the properties of “difficult” sounds like the front fricatives. |
---|---|
ISSN: | 0001-4966 1520-8524 |
DOI: | 10.1121/1.5068354 |