14th International Congress of Phonetic Sciences (ICPhS-14) San Francisco, CA, USA August 1-7, 1999

Speaker Normalization Preserving Regional Accent Differences in Vowel Quality

Patti Adank (1), Vincent J. van Heuven (2), Roeland van Hout (1)

(1) University of Nijmegen/CLS, Nijmegen, The Netherlands
(2) Phonetics Laboratory, Leiden University/HIL, The Netherlands

The ability of several speaker normalization procedures to preserve differences in phonetic quality of vowels from speakers with a different regional accent is evaluated. Also, different ways to represent vowels are compared: F1 vs. F2 in Hz, and F1 vs. F2' in hertz as well as in Bark. A combination of an extrinsic normalization (Gerstman's endpoint normalization) with raw F1 and F2 in hertz appears to be the most suitable approach for acoustic description of regiolectal and/or sociolectal variation in a language.

Full Paper

Bibliographic reference.  Adank, Patti / Heuven, Vincent J. van / Hout, Roeland van (1999): "Speaker normalization preserving regional accent differences in vowel quality", In ICPhS-14, 1593-1596.