Inverse filtering measures for predicting overall voice quality in vowel as well as in continuous speech segments: applications in Praat

Youri Maryn, Paul Corthals, Marc De Bodt, Paul Van Cauwenberge

Acoustic measurement of voice quality, due to several technical challenges associated with extracting F₀ and perturbation measures in continuous speech, traditionally relies on the analysis of a vowel (segment). Clinical and objective voice quality assessment in continuous speech on the other hand is less artificial and more realistic. Acoustic analyses based on inverse filtering, such pitch amplitude (PA) and spectral flatness of the residue signal (SFR) (Davis, 1979), can be used in continuous speech (Parsa & Jamieson, 2001). In this study, the predictive criterion-related validity of inverse filtering measures and autocorrelation analysis with the program Praat was investigated in sustained vowel (3 seconds of /a/) as well as in continuous speech (voiced parts of the first 2 sentences of a standard text) samples provided by 251 subjects with various voice disorders. Furthermore, this validity was also investigated in speech samples containing a concatenation of both. Results, based on correlation statistics, will be discussed in the context of the available literature.

Youri Maryn, MSc
Sint-Jan General Hospital — Department of Otorhinolaryngology and Head & Neck Surgery — Department of Speech-Language Pathology & Audiology — Bruges, Belgium
Paul Corthals, PhD
University College of Ghent — Department of Healthcare ‘Vesalius’ — Ghent, Belgium Universityof Ghent — Department of Speech-Language Pathology & Audiology — Ghent, Belgium
Marc De Bodt, PhD
University Hospital of Antwerp - Department of Otorhinolaryngology and Head & Neck Surgery — Department of Communication Disorders — Antwerp, Belgium
Paul Van Cauwenberge, PhD
University Hospital of Ghent - Department of Otorhinolaryngology and Head & Neck Surgery — Ghent, Belgium

Correspondence: Youri Maryn
Sint-Jan General Hospital
Department of Speech-Language Pathology & Audiology
Ruddershove 10
8000 Bruges
Belgium
Tel.: +32 50 452448 Fax: +32 50 452290
E-mail: youri.maryn@azbrugge.be