A comparison of vowel normalization procedures for language variation research

Abstract

An evaluation of vowel normalization procedures for the purpose of studying language variation is presented. The procedures were compared on how effectively they ͑a͒ preserve phonemic information, ͑b͒ preserve information about the talker's regional background ͑or sociolinguistic information͒, and ͑c͒ minimize anatomical/physiological variation in acoustic representations of vowels. Recordings were made for 80 female talkers and 80 male talkers of Dutch. These talkers were stratified according to their gender and regional background. The normalization procedures were applied to measurements of the fundamental frequency and the first three formant frequencies for a large set of vowel tokens. The normalization procedures were evaluated through statistical pattern analysis. The results show that normalization procedures that use information across multiple vowels ͑''vowel-extrinsic'' information͒ to normalize a single vowel token performed better than those that include only information contained in the vowel token itself ͑''vowel-intrinsic'' information͒. Furthermore, the results show that normalization procedures that operate on individual formants performed better than those that use information across multiple formants ͑e.g., ''formant-extrinsic'' F2-F1).

Dialect divergence and convergence in New Zealand English

“…In order to minimize physiological differences between participants so that the analysis can focus on speech production differences in dialect and style, it is necessary to normalize the vowel formants. As per Adank, Smits, and van Hout (2004), the Lobanov normalization method was used (Lobanov 1971).…”

The effect of L1 regional variation on the perception and production of standard L1 and L2 vowels

“…Dutch production task Figure 4 presents the spectral properties in terms of F1 and F2 of the Dutch vowels produced by the Ghent and Antwerp speakers. All vowels were normalized using the Lobanov normalization method (Lobanov 1971), since it has been shown that this method is effective in eliminating physiological sex-related differences, while preserving regional differences (see Verhoeven and Van Bael 2002a, 2002band Adank et al 2004b). …”

Evaluating vowel pronunciation quality: Formant space matching versus ASR confidence scoring

“…However, it is important to consider the fact that formant locations are influenced not only by the phonemic quality but also by the anatomical/physiological characteristics of the speaker such as vocal tract length and shape. Some vowel normalization methods are effective in reducing speaker differences in terms of anatomical/physiological variations while largely preserving phonemic and socio-linguistic variations [4]. This property of the normalization procedures is instrumental in our native -non-native classification.…”

“…Vowel-extrinsic normalization procedures use the knowledge of the formants of all the vowels of the speaker. In research investigating language variation [4], vowelextrinsic formant-intrinsic normalization procedures have been found effective in reducing differences due to physiological variations while at the same time preserving phonemic and socio-linguistic variation.…”

“…The two methods of normalization that we have explored are described below. 1) Lobanov Normalization: Lobanov's (1971) normalization procedure [4,5,6] standardizes the mean and the standard deviation for each speaker's vowels with the equation:…”