A comparison of vowel normalization procedures for language variation research

Area of Science:

• Linguistics
• Acoustic Phonetics
• Sociolinguistics

Background:

• Studying language variation requires accurate acoustic analysis of speech.
• Vowel normalization is crucial for minimizing non-linguistic variations in acoustic signals.
• Existing normalization methods vary in their effectiveness.

Purpose of the Study:

• To evaluate different vowel normalization procedures for studying language variation.
• To assess the effectiveness of procedures in preserving phonemic and sociolinguistic information.
• To determine which methods best minimize anatomical/physiological variations in vowel acoustics.

Main Methods:

• Compared vowel normalization procedures on Dutch speech data from 160 speakers (80 female, 80 male).
• Stratified speakers by gender and regional background.
• Applied normalization to fundamental frequency and first three formant frequencies.
• Evaluated procedures using statistical pattern analysis.

Main Results:

• Vowel-extrinsic normalization procedures (using multi-vowel information) outperformed vowel-intrinsic methods.
• Procedures operating on individual formants were more effective than those using multi-formant information (e.g., F2-F1).
• Normalization effectiveness was assessed based on preservation of phonemic, sociolinguistic, and anatomical information.

Conclusions:

• Vowel-extrinsic and individual formant normalization strategies are superior for analyzing language variation.
• These methods enhance the reliability of acoustic analysis in sociolinguistic research.
• The findings provide guidance for selecting appropriate normalization techniques in phonetic studies.