Given the acoustic consequences of physiological differences between talkers, there is a practical need for effective and theoretically motivated procedures of vowel normalization to facilitate comparison of speech produced by people who differ by dialect or language. In addition, there is a question whether listeners might utilize a normalization procedure during speech perception. This paper reports the results of two studies that explore these questions—with particular focus on vocal tract length normalization. Drawing on research in speech engineering, where accurate estimates of vocal tract length are needed in some approaches to automatic speech recognition and speaker verification, a new model of vowel normalization is introduced. The model uses a direct measure of average formant spacing (the ΔF) which can be used to measure vocal tract length. The acoustic consequences of vocal tract length differences are removed from vowel measurements by scaling vowel formant measurements by ΔF. Study 1 found that this method is comparable to Nearey’s (1978) uniform normalization method, while providing an explicit vocal tract length interpretation, and a rationalized unit of measure. Study 2 found that uniform normalization measures (which let each formant serve as a noisy estimator of ΔF) improve vowel classification even with only a couple of randomly selected vowel tokens. This suggests that vocal tract length normalization could be involved in speech perception. The ΔF method of vocal tract length normalization for vowels | Laboratory Phonology --> --> Skip to main content Journal article The ΔF method of vocal tract length normalization for vowels Journal article The ΔF method of vocal tract length normalization for vowels Abstract Given the acoustic consequences of physiological differences between talkers, there is a practical need for effective and theoretically motivated procedures of vowel normalization to facilitate comparison of speech produced by people who differ by dialect or language. In addition, there is a question whether listeners might utilize a normalization procedure during speech perception. This paper reports the results of two studies that explore these questions—with particular focus on vocal tract length normalization. Drawing on research in speech engineering, where accurate estimates of vocal tract length are needed in some approaches to automatic speech recognition and speaker verification, a new model of vowel normalization is introduced. The model uses a direct measure of average formant spacing (the ΔF) which can be used to measure vocal tract length. The acoustic consequences of vocal tract length differences are removed from vowel measurements by scaling vowel formant measurements by ΔF. In two studies, the paper will evaluate ΔF vowel normalization as a practical technique for acoustic analysis of language data, and also gauge the feasibility of vocal tract length normalization as a perceptual mechanism. 1.1. Vocal tract length According to the acoustic theory of speech production ( Fant, 1960 ) it is theoretically possible to remove vocal tract length differences from our descriptions of speech acoustics ( Nordström & Lindblom, 1975 ). Once we have measurements of the vocal tract resonant frequencies (F1-F4) we can use the talker-specific vowel formant distributions to estimate normalization factors. It has been argued that perceptual compensation for talker differences must be more than simply a vocal tract length normalization, because acoustic differences between men and women (who tend to differ in vocal tract length) is language-specific; male/female differences depend in part on the language or dialect that they speak ( Johnson, 2005 ). This language-specificity of gender differences suggests that there is more involved than just vocal tract length difference. Instead, there appears to be a performative aspect of gender that is overlaid on physical sex differences in vocal tract length. In vocal tract length normalization ( Nordström & Lindblom, 1975 ), a single normalization scale factor is derived from an estimate of the length of the speaker’s vocal tract. An acoustic factor related to vocal tract length is ϕ, the fundamental frequency of vocal tract resonances in an unconstricted vocal tract ( Paige & Zue, 1970 ; Kirlin, 1978 ; Lammert & Narayanan, 2015 ). The factor ϕ is equal to F1 and the other formants are on odd harmonics of this value: F2 = 3ϕ, F3 = 5ϕ, F4 = 7ϕ. ΔF is simply 2ϕ, and is an estimate of formant spacing in an unconstricted vocal tract ( Reby & McComb, 2003 ), where F1 = ½*ΔF; F2 = 1½*ΔF; F3 = 2½*ΔF, etc. The aims of the paper are to evaluate ΔF vowel normalization as a practical technique for acoustic analysis of language data, and also to gauge the feasibility of vocal tract length normalization as a perceptual mechanism. 2. Methods for vocal tract length normalization The earliest use of vocal tract length as a normalization factor was by Nordström and Lindblom ( 1975 ). They calculated the average third formant (F3) frequency in open vowels (where the frequency of F3 is easily distinguished from F2) and used this to estimate the talker’s vocal tract length. Vocal tract length was then used to scale a talker’s vowel formant measurements onto a ‘standard’ (i.e., male) vocal tract. We can avoid the male-centric bias of this approach by using a talker-independent measurement scale—an estimate of formant spacing in an unconstricted vocal tract, the ΔF. Speech engineers have devised measures of vocal tract length from vowel acoustics ( Paige & Zue, 1970 ; Wakita, 1977 ; Kirlin, 1978 ) for use in automatic speech recognition and speaker identification ( Eide & Gish, 1996 ; Lee & Rose, 1998 ). For example, Kirlin ( 1978 ) used information from the four lowest formants and estimates of formant variances to weight the contributions of formants.

Estimating the vocal tract length (VTL), given the acoustic signal of a vowel sound, is an important problem, which is useful in speaker normalization for vowel recognition, in the inversion problem and in acoustic-phonetic studies. The common approach of using the formant data to estimate VTL works for a neutral vowel approximating a uniform tube. However, for natural vowels, formant data shift considerably away from the resonant frequencies of a uniform tube. The proposed method is motivated from these observations: (a) the frequency of a spectral valley, F v , depends inversely on VTL; (b) there is much smaller shift in F v , across vowels, from the corresponding valley frequency of a uniform tube; (c) F v can be estimated from the spectral envelope itself. VTL has been estimated for the Pe-terson and Barney (33 male and 28 female speakers) and the TIMIT (326 male and 136 female speakers) databases. When the estimated F v is used for normalization, the spread in the formant data due to gender differences is considerably reduced. The normalization procedure is vowel and speaker intrinsic. Additionally, we report applications such as front/back classi-fication, gender recognition and phonetic feature mapping. ISCA Archive - Estimation of the Vocal Tract Length of Vowel Sounds Based on the Frequency of the Significant Spectral Valley ISCA Archive Interspeech 2018 ISCA Archive Interspeech 2018 Estimation of the Vocal Tract Length of Vowel Sounds Based on the Frequency of the Significant Spectral Valley TV Ananthapadmanabha, Ramakrishnan A G Estimating the vocal tract length (VTL), given the acoustic signal of a vowel sound, is an important problem, which is useful in speaker normalization for vowel recognition, in the inversion problem and in acoustic-phonetic studies. The common approach of using the formant data to estimate VTL works for a neutral vowel approximating a uniform tube. However, for natural vowels, formant data shift considerably away from the resonant frequencies of a uniform tube. The proposed method is motivated from these observations: (a) the frequency of a spectral valley, F_v, depends inversely on VTL; (b) there is much smaller shift in F_v, across vowels, from the corresponding valley frequency of a uniform tube; (c) F_v can be estimated from the spectral envelope itself. VTL has been estimated for the Peterson and Barney (33 male and 28 female speakers) and the TIMIT (326 male and 136 female speakers) databases. When the estimated F_v is used for normalization, the spread in the formant data due to gender differences is considerably reduced. The normalization procedure is vowel and speaker intrinsic. Additionally, we report applications such as Front/Back classification, gender recognition and phonetic feature mapping. @inproceedings{ananthapadmanabha18_interspeech, title = {{Estimation of the Vocal Tract Length of Vowel Sounds Based on the Frequency of the Significant Spectral Valley}}, author = {TV Ananthapadmanabha and Ramakrishnan {A G}}, year = {2018}, booktitle = {{Interspeech 2018}}, pages = {2102--2106}, doi = {10.21437/Interspeech.2018-1105}, issn = {2958-1796}, } Cite as : Ananthapadmanabha, T., A G, R.

Relationships between human voice qualities and body build are fairly well researched. However, a gap in the knowledge remains about voice and body composition associations. This study aims to assess the relationships between voice parameters and body composition (fat mass, FM and fat-free mass, FFM) of adults. A total of 204 participants (of which 81 are men) aged 18-72 years (mean: 34.1 ± 13.6 y.) took part in the examination. All of them had four tasks to complete: (i) preliminary survey, (ii) voice recordings with vowels and sentence stimuli, (iii) anthropometric measurements, and (iv) body composition analysis using bioimpedance to describe fat (FM) and lean (FFM) body mass. Canonical correlation analysis (CCA) and partial correlations controlling for age were applied. Reverse correlations between mean pitch (F_o) and shoulder circumference and averaged FFM of arms and positive relations between the first formant (F₁) and FM, BMI and shoulder-to-hip ratio were found in the male group. In the female group, positive associations between voice intensity and BMI, waist circumference, FM of trunk and waist-to-hip ratio and a relation of the same direction between maximum phonation time and FFM were found. These findings may be important for forensic science professionals and laryngologists, but further studies are essential.