A comparison of three speaker-intrinsic vowel formant frequency normalization algorithms for sociophonetics

Abstract

This article evaluates a speaker-intrinsic vowel formant frequency normalization algorithm initially proposed in Watt & Fabricius (2002). We compare how well this routine, known as the S-centroid procedure, performs as a sociophonetic research tool in three ways: reducing variance in area ratios of vowel spaces (by attempting to equalize vowel space areas); improving overlap of vowel polygons; and reproducing relative positions of vowel means within the vowel space, compared with formant data in raw Hertz. The study uses existing data sets of vowel formant data from two varieties of English, Received Pronunciation and Aberdeen English (northeast Scotland). We conclude that, for the data examined here, the S-centroid W&F procedure performs at least as well as the two speakerintrinsic, vowel-extrinsic, formant-intrinsic normalization methods rated as best performing by Adank (2003): Lobanov's (1971) z-score procedure and Nearey's (1978) individual log-mean procedure (CLIH i4 in Adank [2003], CLIH i2 as tested here), and in some test cases better than the latter.The S-centroid vowel normalization procedure, originally presented in detail in Watt & Fabricius (2002), was developed to further research on variation and change in British English vowel systems (e.g., Watt & Tillotson's [2001] discussion of Bradford English). It offered a new normalization method tailored specifically for sociophonetics, optimizing as far as possible comparisons of vowel systems from different speakers, both male and female, without priorThe authors thank Caroline Moreiras and Bronwen Evans for access to Moreiras ' (2006) unpublished vowel formant data, Jillian Oddie for assisting with the Aberdeen data collection and analysis, Tyler Kendall for advice on the NORM suite, Victoria Watt with help with processing the Aberdeen data, and Bernhard Fabricius for assistance with exploratory mathematical and programming tasks. Data collection in Aberdeen was

Accommodation or political identity: Scottish members of the UK Parliament

“…i In addition to CAT, tokens of the Scottish MEET vowel (akin to Wells' FLEECE set) were coded for purposes of vowel normalization (Fabricius et al 2009). We also coded for STRUT to enable the speakerspecific comparison in the centrality of the CAT vowel.…”

Intrinsic-cum-extrinsic normalization of formant data of vowels

“…Hence the procedure has to be strictly called "speaker-intrinsic, formant-extrinsic, and vowelintrinsic" normalization. 5 Instead, for the sake of brevity, we refer to the procedure as intrinsic normalization.…”

“…Hence, the statistics of the formant data across all vowels, instead of vowel specific statistics, are used in the existing extrinsic procedures. [4][5][6]11 However, we make use of vowel specific statistics, the mean l(i, j), and the standard deviation r(i, j) of vowel j. During the process of the proposed extrinsic normalization, the identity of the vowel sample is also determined.…”

“…The effectiveness of some select vowel normalization methods have been compared based on certain objective criteria. [5][6][7] Carpenter and Govindarajan 8 give a brief description as well as an evaluation of 32 intrinsic and 128 extrinsic procedures for the vowel classification task. Normalization in the context of sociolinguistics has also been reported.…”

“…The performance of the proposed procedure compares well with, or is better than, two top-ranked normalization procedures. 4,5 2. Proposed method…”