Formant Measurements Research Articles

Changes to articulation following LSVT® and traditional dysarthria therapy in non-progressive dysarthria

The present study aimed to evaluate the effects of the Lee Silverman Voice Treatment (LSVT®) on acoustic and perceptual measures of articulation in non-progressive dysarthria in comparison to traditional dysarthria therapy. The study involved 26 individuals with non-progressive dysarthria who were randomly allocated to receive either LSVT® or traditional dysarthria therapy (TRAD), both of which were administered for 16 hourly sessions over 4 weeks. Participants' speech samples were collected over a total of six testing sessions during three assessment phases: (1) prior to treatment, (2) immediately post-treatment, and (3) 6 months post-treatment (FU). Speech samples were analysed perceptually to determine articulatory precision and intelligibility as well as acoustically using vowel space (and vowel formant measures) and first moment differences. Results revealed short and long-term significant increases in vowel space area following LSVT®. Significantly increased intelligibility was also found at FU in the LSVT® group. No significant differences between groups for any variables were found. The study reveals that LSVT® may be a suitable treatment option for improving vowel articulation and subsequent intelligibility in some individuals with non-progressive dysarthria.

Formant frequencies of vowels in 13 accents of the British Isles

This study is a formant-based investigation of the vowels of male speakers in 13 accents of the British Isles. It provides F1/F2 graphs (obtained with a semi-automatic method) which could be used as starting points for more thorough analyses. The article focuses on both phonetic realization and systemic phenomena, and it also provides detailed information on automatic formant measurements. The aim is to obtain an up-to-date picture of within- and between-accent vowel variation in the British Isles. F1/F2 graphs plot z-scored Bark-transformed formant frequencies, and values in Hertz are also provided. Along with the findings, a number of methodological issues are addressed.

The effect of increased loudness on anticipatory coarticulation of steady-state formants in a vowel-consonant-vowel context for patients with Parkinson’s disease.

The current investigation examines the effect of increased loudness on anticipatory coarticulation of the vowels [i, ɑ, u] on the preceding schwa vowel for patients with Parkinson’s disease. Four speakers, three males and one female with speech severity levels between 8% and 49%, are involved in pre-treatment, treatment, and post-treatment conditions. During treatment, participants use an auditory device which plays speech-babble into one ear while they are speaking. The speech-babble competition results in an automatic loudness increase (Lombard effect) in the talker wearing the device. Participants produced target nonsense CVC words embedded in carrier sentences at comfortable loudness levels and at an average of 3 dB above that level. The steady-state F1, F2, and F3 formant frequencies of the preceding [ə] were analyzed to determine whether increased loudness results in increased anticipatory coarticulation from the [i, ɑ, u] vowels in the CVC words, as treatment progresses. Results using steady-state formant frequency measures from this study will be compared with previous studies that used F2 trajectory measures [Tjaden and Sussman (2006); Weismer et al. (1995); Tjaden (2003)]. (Work supported by NIH 1R01DC009409-01)

An acoustic and electroglottographic study of phonation contrast in Yi.

This study examines the phonation contrast of Yi, a Tibeto-Burman language of Southwestern China with phonation as a phonemic dimension (traditionally described as tense versus lax contrast). The language has seven monophthongs as well as three tones (high, mid, and low). Tense versus lax phonation contrast can apply to all the vowels and two tones (mid and low) in the language. In this study, both acoustic and electroglottographic data were collected with minimal pairs for all combinations of tones and vowels. Extensive acoustic measures were done by VOICESAUCE, including F0, H1, H2, H1-H2, H2-H4, H1-A1, H1-A2, F1, F2, and cepstral peak prominence (CPP); EGG measures were closed quotient (CQ) and peak-velocity (PV). Preliminary data suggest that H1-H2, CPP, CQ, and PV measures can exclusively distinguish the two phonation types. Moreover, EGG data also suggest voice quality interacts with tones, but the patterns vary among speakers. This might due to the different strategies used by speakers to produce the tense versus lax contrast. In addition, formant frequency measures also show salient differences between tense and lax vowels; tense vowels generally have higher F1 than the lax counterparts. All of the facts suggest that phonation contrasts in Yi are realized across different phonetic dimensions.

Computational Methods for Normalizing Acoustic Vowel Data for Talker Differences

AbstractDirect acoustic–phonetic comparison of vowels produced by different talkers can be difficult due to the effects of anatomical and physiological talker differences on first and second formant frequencies. Since the mid‐1960s, numerous computational algorithms have been proposed to normalize formant frequency measures and reduce the effects of talker‐specific variability. These algorithms vary widely in their conceptualization: some require samples of the entire vowel space (vowel extrinsic), whereas others can be conducted on a single vowel class (vowel intrinsic); some are based on a single‐scale factor to align broad classes of talkers (such as females and males), whereas others use scale factors and/or vowel space ranges for individual talkers to align their vowel systems in the same acoustic–phonetic space. In this study, an overview of the most commonly used normalization algorithms is provided and the advantages and disadvantages of each of the approaches are discussed.

An acoustical comparison of English tense and lax vowels.

Several studies on the pronunciation of English vowels point out that Korean learners have difficulty producing English tense and lax vowel pairs. The acoustic comparisons of those studies are mostly based on the formant measurement at one time point of a given vowel section. However, the English lax vowels usually show dynamic changes across their syllable peaks and subjects’ English levels account for various conflicting results. The purposes of this paper are to compare the temporal duration and dynamic formant tracks of English tense and lax vowel pairs produced by five Korean and five American males. Results showed that both the Korean and American males produced the vowels with comparable durations. The duration of the front tense‐lax vowel pair was longer than that of the back vowel pair. From the formant track comparisons, the American males produced the tense and lax pairs much more distinctly than the Korean males. The results suggest that the Korean males should pay attention to the F1 and F2 movements, i.e., the jaw and tongue movements, in order to match those of the American males. Further studies are recommended on the auditorily acceptable ranges of F2 variation for the lax vowels.

Effect of word length on vowel production by Mandarin and American speakers: Comparison of [i] and [ɪ

Yang (2008) found that Mandarin speakers’ productions of [eɪ] and [ε] (as in gate‐get) showed much less difference than Americans made. In addition, they made a smaller distinction in disyllables than in monosyllables. The current study compares the production of [i] and [ɪ] (beat‐bit) in the same conditions. Target‐vowels were embedded in Stop‐V‐Stop context in carrier sentences. Vowel durations were measured after stop release until final closure. Formant measurements were made at two time‐points for each token. Preliminary analysis extends previous results of reduced contrast between tense and lax vowels by the Mandarin speakers to vowels [i]‐[ɪ] and also shows greater reduction of the contrast in disyllabic words relative to monosyllables. The Mandarin speakers’ formants for [i]‐[ɪ] were almost identical in the disyllabic words, and their duration ratios were not significantly different, whereas American speakers’ measurements were very different between two vowels and the same in disyllables as monosyllables. The current study provides further evidence for an effect of word length on English vowel contrasts by Mandarin speakers.

An algorithm for token-by-token linear predictive coding analysis of oral monophthongs.

Linear Predictive Coding (LPC) is the most reliable vowel formant estimation method. LPC trajectories closely approximate formants and are preferable as a means of avoiding human bias. Nevertheless, formant analysis by LPC is not error free, as it can be affected by fundamental frequency, noise and vocal tract morphology. Error minimization is usually achieved through manual checking of LPC results with spectral contour. An algorithm to find optimal LPC parameters (Burg method) on a token by token basis was developed and implemented as a Praat script. The algorithm minimizes inter- and intra-vowel formant variability without knowledge of vowel identity. Algorithm results were compared with results of manual adjustment of LPC parameters on a trial and error basis on a corpus of 180 stressed and unstressed oral monophthongs of Brazilian Portuguese from a male speaker. First and second formants linear correlations are r2 = 0.98 and 0.99 respectively. The areas of /i, a, u/ vocalic triangles (mean values as vertices) of manual and automatic measurements showed only 4.5% difference. These results suggest the algorithm can be generalized to any speaker and speed up the measurement of vowel formants in large corpora. [Work supported by CNPq and Fapesp.]

Développement de la parole et émergence de la structure prosodique chez l’enfant : une étude de l’accent d’emphase en français

AbstractThis article presents the results of a study on articulatory and acoustical correlates of contrastive focus in French in five children aged 4 and 5 and five adults. The speakers repeated the sequence [baba] in two prosodic contexts: neutral conditions and contrastive focus. The acoustic signal and the trajectories of three sensors placed on the subjects’ upper lip, lower lip, and chin were recorded using an Optotrak system. Articulatory movements were analyzed for the two syllables in the sequence, in each of the two prosodic conditions. Formant measurements, sound intensity, fundamental frequency, and acoustic duration of the segments were also extracted from the acoustic signal. The results show that the effects of contrastive focus are smaller in the children than the adult speakers. The results are interpreted in light of recent theories on the emergence of spoken language in children.

The Development of Distinct Speaking Styles in Preschool Children

To examine when and how socially conditioned distinct speaking styles emerge in typically developing preschool children's speech. Thirty preschool children, ages 3, 4, and 5 years old, produced target monosyllabic words with monophthongal vowels in different social-functional contexts designed to elicit clear and casual speaking styles. Thirty adult listeners were used to assess whether and at what age style differences were perceptible. Children's speech was acoustically analyzed to evaluate how style-dependent differences were produced. The ratings indicated that listeners could not discern style differences in 3-year-olds' speech but could hear distinct styles in 4-year-olds' and especially in 5-year-olds' speech. The acoustic measurements were consistent with these results: Style-dependent differences in 4- and 5-year-olds' words included shorter vowel durations and lower fundamental frequency in clear compared with casual speech words. Five-year-olds' clear speech words also had more final stop releases and initial sibilants with higher spectral energy than did their casual speech words. Formant frequency measures showed no style-dependent differences in vowel production at any age nor any differences in initial stop voice onset times. Overall, the findings suggest that distinct styles develop slowly and that early style-dependent differences in children's speech are unlike those observed in adult clear and casual speech. Children may not develop adultlike styles until they have acquired expert articulatory control and the ability to highlight the internal structure of an articulatory plan for a listener.

Acoustic correlates of speaker discrimination in English.

This study investigated speaker discrimination in utterances varying in syllable length and speaker gender taken from the TIMIT corpus of American English. Twenty native English speakers presented one‐, two‐, and three‐syllable utterances (within speaker gender) in a two‐alternative forced‐choice task. Perception results were analyzed in light of both source level (F0 and long‐term average spectrum of LPC residuals) and formant level measurements (F1–F4). Results showed that male speakers were discriminated better than female speakers. Source features (F0 and LTAS of LPC residuals) significantly predicted listener response, while higher spectral information (F1–F4) had little effect. The varying importance of vocal source and vocal tract characteristics in speaker discrimination is discussed.

Forensic Phonetics

AbstractAn overview of forensic phonetics is presented, focusing on speaker identification as its core task. Speaker profiling/speaker classification is applied when the offender has been recorded, but no suspect has been found. Auditory speaker identification by victims and witnesses becomes relevant when no speech recording of the offender is available. It can take the form of familiar‐speaker identification or unfamiliar‐speaker identification, and in the latter case a voice line‐up/voice parade can be carried out. When recordings of both the offender and a suspect are available, a voice comparison is done by an expert in forensic speech analysis. Current issues and domains in voice comparison analysis include the Bayesian approach to forensic reasoning and the Likelihood Ratio, the use of formant frequency measurements, non‐analytic perception and Exemplar Theory, forensic automatic speaker identification, and the interaction between different methods.

Vowel place detection for a knowledge-based speech recognition system

This work aims to detect vowel place as part of a knowledge-based speech recognition system. Vowel place was classified into 6 groups based on tongue advancement [Front/Back] and height [High/Mid/Low]. Experiments were performed using 300 /hVd/ utterance data from Hillenbrand [J. Acoust. Soc. Am. 97, 3099-3111] and 6600 TIMIT vowels. Features used include fundamental frequency (F0) and formant value (F1̃F3), where formant measurements were classified into separate groups using F0 measurements. The nearest class was found using a simple Mahalanobis distance measure, and yielded a 91.5% classification rate for the /hVd/ data. The results for the TIMIT data were 64.4%, and error analysis with regard to adjacent segment manner and place was carried out to observe the effects of coarticulation, which was not observed in the /hVd/ data.

Production of American English [eI] and [ε] by Mandarin speakers: mono- vs. disyllabic words

Mandarin productions of English tense vs. lax vowels are difficult for English speakers to differentiate (Chen et al. 2006). In this study the production of American English [eI] and [ε] by Mandarin and American participants was investigated. The target vowels were embedded in monosyllabic and disyllabic words with a stop-V-voiceless stop context in carrier sentences of variable length. There were two formant measurements for each vowel (20% after the onset and before the end of the vowel). In the monosyllabic condition, Mandarin speakers' performance was similar to American speakers in duration and formant values except that [eI] started too low. However, in the disyllabic condition, Mandarin speakers' formants for [eI] and [ε] were overlapped and there was no significant durational distinction between [eI] and [ε]. One possible account for this difference is that when producing a disyllabic word, Mandarin speakers tend to shorten the vowel duration of the first syllable to compensate for the production of the following syllable. The difference in syllable conditions may reflect a tendency that Chinese regularizes word durations regardless of the number of syllables. Reference: Chen, Y. et al. (2006). Vowel production by Mandarin speakers of English. Clinical linguistics and phonetics, 15, 427-440

What’s wrong with these formants?

Speech is frequently modeled as an all-pole (autoregressive) linear filter applied to a voice source, commonly called Linear Predictive Coding. This model has been used for many years to provide measurements of the resonance frequencies (formants) characterizing speech sounds. That notwithstanding, it has also been well recognized that the use of this method to measure formants is very often fraught with problems, but just how serious are these problems, and can anything be done to improve upon LPC formant measurements? A few examples will demonstrate that the problems plaguing LPC formant measurements are too often intractable, and frequently yield formant values which are nowhere near the truth. For example, the voice bar will be shown to falsely lower the reported first formant value in many simple cases. It might even be suggested that reliance on this technique has impeded our understanding of exotic speech sounds which do not display a typical formant pattern. Precision formant measurements can be provided in most instances by reassigned spectrograms that have been specially processed to highlight signal components. Examples will be shown and used as benchmarks against which the LPC method can at last be properly evaluated.

An acoustic description of the vowels of northern and southern standard Dutch II: Regional varieties

An analysis is presented of regional variation patterns in the vowel system of Standard Dutch as spoken in the Netherlands (Northern Standard Dutch) and Flanders (Southern Standard Dutch). The speech material consisted of read monosyllabic utterances in a neutral consonantal context (i.e., /sVs/). The analyses were based on measurements of the duration and the frequencies of the first two formants of the vowel tokens. Recordings were made for 80 Dutch and 80 Flemish speakers, who were stratified for the social factors gender and region. These 160 speakers were distributed across four regions in the Netherlands and four regions in Flanders. Differences between regional varieties were found for duration, steady-state formant frequencies, and spectral change of formant frequencies. Variation patterns in the spectral characteristics of the long mid vowels /e o ø/ and the diphthongal vowels /ei oey bacwards c u/ were in accordance with a recent theory of pronunciation change in Standard Dutch. Finally, it was found that regional information was present in the steady-state formant frequency measurements of vowels produced by professional language users.

Spectral Evaluation of Hypernasality in Children with Repaired Cleft Palate

Objective: To identify the characteristics of the spectrum of hypernasality in relation to the formant amplitudes for the vowel ‘a’ (∂ in the International Phonetic Association alphabet chart of vowels) for children with repaired cleft palate and the relationship of hypernasality with other acoustic parameters.Patients and Methods: Fifty three children with repaired cleft palate were divided into 2 groups, based on perceptual speech rating: moderate to severe hypernasality (n = 33) and mild hypernasality (n = 20). The control group comprised 20 children without cleft palate. Sound segments ‘ka’ containing the vowel ‘a’ were recorded, digitised, and analysed to identify formant pattern, breathiness values, and amplitudes for one-third octave band frequency spectra using computer software.Results: Differences in frequency values obtained between the hypernasal groups and normal children for the fundamental frequency (F0) and formants F1 and F2 were not statistically significant. Significant differences in breathiness values were obtained for children with moderate to severe hypernasalality and those with mild hypernasality compared with normal children. Spectral analysis revealed an increase in amplitudes in frequency bands between F1 and F2 and spectral dips at 630 to 800 Hz band and at F2 with an additional Fn peak at 800 to 1000 Hz band in subjects with moderate to severe hypernasality.Conclusions: Formant amplitude measurement using one-third octave frequency band spectral analysis revealed characteristic features for the hypernasal vowel ‘a’ as an increase in amplitudes in some frequency bands and an additional spectral peak Fn between F1 and F2.

Resonance in an exemplar-based lexicon: The emergence of social identity and phonology

Two sets of data are discussed in terms of an exemplar-resonance model of the lexicon. First, a cross-linguistic review of vowel formant measurements indicate that phonetic differences between male and female talkers are a function of language, dissociated to a certain extent from vocal tract length. Second, an auditory word recognition study [Strand (2000). Gender Stereotype Effects in Speech Processing. Ph.D. Dissertation, Ohio State University] indicates that listeners can process words faster when the talker has a stereotypical sounding voice. An exemplar-resonance model of perception derives these effects suggesting that reentrant pathways [Edelman (1987). Neural Darwinism: The theory of neuronal group selection. New York: Basic Books] between cognitive categories and detailed exemplars of them leads to the emergence of social and linguistic entities.

On the number of channels needed to classify vowels: Implications for cochlear implants

In cochlear implants the incoming signal is analyzed by a bank of filters. Each filter is associated with an electrode to constitute a channel. The present research seeks to determine the number of channels needed for optimal vowel classification. Formant measurements of vowels produced by men and women [Hillenbrand et al., J. Acoust. Soc. Am. 97, 3099–3111 (1995)] were converted to channel assignments. The number of channels varied from 4 to 20 over two frequency ranges (180–4000 and 180–6000 Hz) in equal bark steps. Channel assignments were submitted to linear discriminant analysis (LDA). Classification accuracy increased with the number of channels, ranging from 30% with 4 channels to 98% with 20 channels, both for the female voice. To determine asymptotic performance, LDA classification scores were plotted against the number of channels and fitted with quadratic equations. The number of channels at which no further improvement occurred was determined, averaging 19 across all conditions with little variation. This number of channels seems to resolve the frequency range spanned by the first three formants finely enough to maximize vowel classification. This resolution may not be achieved using six or eight channels as previously proposed. [Work supported by NIH.]

LOW BACK VOWEL MERGER IN MISSOURI SPEECH: ACOUSTIC DESCRIPTION AND EXPLANATION

Across much of the United States, the phonemic contrast between the vowels of cot and caught is being lost through a sound change known as the low back vowel merger. This paper examines the spread of this merger in the state of Missouri. Acoustic examination of F0, F1,F2, F3, and vowel duration reveals that speakers in the greater St. Louis area maintain the phonemic distinction between /a/ andɔ/, while in much of the state, this distinction is being lost or diminished. In addition to static formant measures, the formant trajectories of the two vowels are examined, and it is found that in St. Louis speech, the VC consonantal transition of F2 is accomplished more quickly for/ɔ/ than for /a/. Although the F2 transitions of the two vowels differ, their overall spectral shapes are more similar than other comparable vowel pairs such as /æ/ and /g=e/, which are not undergoing widespread merger. The dynamic similarity between /a/ and /ɔ/ is posited as a partial explanation for why this particular merger is spreading so rapidly throughout the United States.