F2 Plane Research Articles

Perceptual Identification of Polish Vowels Due to F0 Changes

The paper investigates the interdependence between the perceptual identification of the vocalic quality of six isolated Polish vowels traditionally defined by the spectral envelope and the fundamental frequency F0. The stimuli used in the listening experiments were natural female and male voices, which were modified by changing the F0 values in the ±1 octave range. The results were then compared with the outcome of the experiments on fully synthetic voices. Despite the differences in the generation of the investigated stimuli and their technical quality, consistent results were obtained. They confirmed the findings that in the perceptual identification of vowels of key importance is not only the position of the formants on the F1 × F2 plane but also their relationship to F0, the connection between the formants and the harmonics and other factors. The paper presents, in quantitative terms, all possible kinds of perceptual shifts of Polish vowels from one phonetic category to another in the function of voice pitch. An additional perceptual experiment was also conducted to check a broader range of F0 changes and their impact on the identification of vowels in CVC (consonant, vowel, consonant) structures. A mismatch between the formants and the glottal tone value can lead to a change in phonetic category.

Koyee Vowels: An Acoustic Analysis

This paper focuses on the acoustic analysis of Koyee basic vowels. Koyee exhibits basically six vowels. In the acoustic analysis of the basic vowel sounds in the Koyee language, the first formant (F2) are identified as the acoustic analysis cues and they are measured both male and female Koyee speakers individually and in average. The average of F1 and F2 was plotted in F1 versus F2 plane and later compared them plotting them in the same acoustic vowel space. The study shows that female acoustic vowel space is slightly larger than the male. This is because of the differences of the vocal tract of female and male. Although the exact formant frequency values of the vowels produced by male and female speakers varied, the ratio of the difference of the formant frequency values of the different vowels is similar for both male and female speakers. Further, the female voice is more conservative than those of the male speakers because the males are very often in contact with other language speakers.

Contribution of formant frequency information to vowel perception in steady-state noise by cochlear implant users

Cochlear implant (CI) recipients have difficulty understanding speech in noise even at moderate signal-to-noise ratios. Knowing the mechanisms they use to understand speech in noise may facilitate the search for better speech processing algorithms. In the present study, a computational model is used to assess whether CI users' vowel identification in noise can be explained by formant frequency cues (F1 and F2). Vowel identification was tested with 12 unilateral CI users in quiet and in noise. Formant cues were measured from vowels in each condition, specific to each subject's speech processor. Noise distorted the location of vowels in the F2 vs F1 plane in comparison to quiet. The best fit model to subjects' data in quiet produced model predictions in noise that were within 8% of actual scores on average. Predictions in noise were much better when assuming that subjects used a priori knowledge regarding how formant information is degraded in noise (experiment 1). However, the model's best fit to subjects' confusion matrices in noise was worse than in quiet, suggesting that CI users utilize formant cues to identify vowels in noise, but to a different extent than how they identify vowels in quiet (experiment 2).

Vowel density differences among regional dialects of American English.

Over the past decade there has been considerable interest in differences in the size and shape of vowel spaces across different ages and genders of speakers and across different regional dialects of English. Vowel spaces provide information about the distribution of vowels in the F1 by F2 plane in an individual speaker or group of speakers. Recent studies have investigated which vowels should be used to establish the spatial boundaries and what criteria should be used in drawing the boundaries. However, the weakness of all such approaches is that they do not take into account the number of vowels produced in different portions of the vowel space (the occurrence of a single vowel in a peripheral location is sufficient to enlarge the vowel space). This study proposes to expand and refine the concept of vowel space by creating vowel density maps across different regional dialects of American English. These maps represent a three‐dimensional overlay of relative vowel frequencies onto the F1 × F2 plane (the borders of which represent the vowel space). Dialectal differences will be described in terms of differences in the intersection of vowel space areas and differences in vowel density within different portions of the intersected spaces.

Vowel spaces in Swedish adolescents with cochlear implants

This paper examines vowel production in Swedish adolescents with cochlear implants. Twelve adolescents with cochlear implants and 11 adolescents with normal hearing participated. Measurements were made of the first and second formants in all the nine long Swedish vowels. The values in hertz were bark-transformed, and two measures of the size of the vowel space were obtained. The first of them was the average Euclidean distance in the F1-F2 plane between the nine vowels and the mean F1 and F2 values of all the vowels. The second was the mean Euclidean distance in the F1-F2 plane between all the vowels. The results showed a significant difference for both vowel space measures between the two groups of adolescents. The cochlear implant users had a smaller space than the adolescents with normal hearing. In general, the size of the vowel space showed no correlations with measures of receptive and productive linguistic abilities. However, the results of an identification test showed that the listeners made more confusions of the vowels produced by speakers who had a small mean distance in the F1-F2 plane between all the vowels.

Dialectal differences in dynamic formant patterns in vowels.

This study examines basic acoustic variation in the dynamic patterns of vowel formants among three regional variants of American English spoken in southeastern Wisconsin (affected by the Northern Cities Shift), western North Carolina (Appalachian English affected by the Southern Vowel Shift), and central Ohio (not considered to be affected currently by any vowel shift). Three groups of speakers (including men and women) produced vowels in citation form (in a /hVd/ context) and in sentences. The sentence material elicited two degrees of vowel emphasis (high and low) in the /bVd/ and /bVt/ contexts. The frequencies and amplitudes of the first three formants were extracted at points corresponding to 20%‐35%‐50%‐65%‐80% of the vowel’s duration. A set of dynamic measures was then calculated using these values that included overall signed/unsigned change in formant frequencies, vector and trajectory lengths, direction/angle of vowel movement in the F1 by F2 plane, and rate of frequency change. The results show significant cross‐dialectal differences in formant patterns in monophthongs, phonemic diphthongs, and nonphonemic diphthongs. The nature and extent of these formant changes varied as a function of speaker dialect in a manner apart from expected variation resulting from phonetic and prosodic context and speaking style. [Work supported by NIH.]

Vowel production in Swedish children with cochlear implants

This study analyzed vowel production of Swedish children with cochlear implants. Swedish has a rich vowel system with nine vowels. Minimal word pairs were used as linguistic material. The age of the children varied between 6 and 13 years. The age at implantation was 1.2–10 years. The first two formants were measured for each vowel. To estimate the vowel space, the average Euclidean distance in the F1–F2 plane between each vowel and the mean first and second formant frequencies of all the vowels was calculated for each child. Preliminary results from 15 children with cochlear implants and 11 control subjects with normal hearing show no difference between the two groups, due to a large amount of variability in the group with cochlear implants. The relationship between the vowel space and the speech perception of each child with a cochlear implant was analyzed using correlations. Results show no strong correlation between the two measures, again due to the variability within the group. [Work supported by the Sunnderdahl Foundation.]

Testing theories of vowel inherent spectral change

Models which include information about vowel inherent spectral change (VISC) have higher correct classification rates, and greater correlation with human response patterns, compared to models which only include static spectral information. Three hypotheses have been proposed as to the aspect of VISC which is relevant for human perception. All three agree that the initial target (e.g., F1 and F2 values at the beginning of the vowel) is important. The dual-target hypothesis states that the final target (e.g., F1 and F2 values at the end of the vowel) is the other relevant factor. The target-plus-slope hypothesis states that it is the rate of change in spectral values over time that is relevant. The target-plus-direction hypothesis states that only the direction of formant movement (e.g., in an F1–F2 plane) is relevant. Synthetic /e/−/I /−/ε/ stimuli with a fixed initial target but varying final target, slope, direction, and duration were identified by monolingual Western Canadian English listeners. Models of the listeners perception were parametrized according to the three hypotheses. Results indicate that the final−target model outperformed the slope modal. Neither the linear−final−target model nor the direction model clearly outperformed the other, but a final−target model including a cubic term was superior.

Neandertal vocal tract

Potential speech abilities constitute a key component in the description of the Neandertals and their relations with modern Homo Sapiens. Since Lieberman & Crelin postulated in 1971 the theory that “Neanderthal man did not have the anatomical prerequisites for producing the full range of human speech” their speech capability has been a subject of hot debate for over 30 years, and remains a controversial question. In this study, we first question the methodology adopted by Lieberman and Crelin, and we point out articulatory and acoustic flaws in the data and the modeling. Then we propose a general articulatory-acoustic framework for testing the acoustic consequences of the trade-off between oral and pharyngeal cavities. Specifically, following Honda & Tiede (1998), we characterize this trade-off by a Laryngeal Height Index (LHI) corresponding to the length ratio of the pharyngeal cavity to the oral cavity. Using an anthropomorphic articulatory model controlled by lips, jaw, tongue and larynx parameters, we can generate the Maximal Vowel Space (MVS), which is a triangle in the F1 / F2 plane, the three point vowels /a/, /i/, and /u/ being located at its three extremities. We sample the evolution of the position of the larynx from birth to adulthood with four different LHI values, and we show that the associated MVS are very similar. Therefore, the MVS of a given vocal tract does not depend on the LHI: gestures of the tongue body, lips and jaw allow compensations for differences in the ratio between the dimensions of the oral cavity and pharynx. We then infer that the vowel space of Neandertals (with high or low larynx) was potentially no smaller than that of a modern human and that Neandertals could produce all the vowels of the world’s languages. Neandertals were no more vocally handicapped than children at birth are. Therefore, there is no reason to believe that the lowering of the larynx and a concomitant increase in pharynx size are necessary evolutionary pre-adaptations for speech. However, since our study is strictly limited to the morphological and acoustic aspects of the vocal tract, we cannot offer any definitive answer to the question of whether Neandertals could produce human speech or not.

The effect of masking noise on acoustic-phonetic contrasts in post-lingually deafened cochlear implant users

This study examined the effect on the vowel contrast distance (average inter-vowel distance in the F1−F2 plane) of gradually decreasing the signal-to-noise ratio (SNR) in the auditory feedback of a post-lingually deafened cochlear implant (CI) user at 1-month and 1-year following CI processor activation. Masking noise, mixed with normal levels of speech feedback, was presented through the headpiece of a research sound processor to the CI user. As a control, an analogous procedure was used for a normal-hearing speaker where the masking noise and speech feedback were delivered over headphones. The SNR was gradually decreased over seven steps as the speakers produced ten repetitions of two vowel contrasts (æ\/[g\/] and i\/u). Speech SPL and vowel contrast distance were measured at all seven masking noise levels. Data from both subjects showed that SPL gradually increased with decreased SNR, while contrast distance decreased. The effect was greater after 1 year of experience with a CI than at 1 month. The effect in the NH speaker was similar to that noted in the CI user after 1 year of experience. Data from additional subjects will be analyzed and reported. [Work supported by NIH Grant No. R01 DC03007.]

Amplitude and phase of distortion product otoacoustic emissions in the guinea pig in an (f1 ,f2) area study.

Lower sideband distortion product otoacoustic emissions (DPOAEs), measured in the ear canal upon stimulation with two continuous pure tones, are the result of interfering contributions from two different mechanisms, the nonlinear distortion component and the linear reflection component. The two contributors have been shown to have a different amplitude and, in particular, a different phase behavior as a function of the stimulus frequencies. The dominance of either component was investigated in an extensive (f1 ,f2) area study of DPOAE amplitude and phase in the guinea pig, which allows for both qualitative and quantitative analysis of isophase contours. Making a minimum of additional assumptions, simple relations between the direction of constant phase in the (f ,f2) plane and the group delays in f1-sweep, f2-sweep, and fixed f2/f1 paradigms can be derived, both for distortion (wave-fixed) and reflection (place-fixed) components. The experimental data indicate the presence of both components in the lower sideband DPOAEs, with the reflection component as the dominant contributor for low f2/f1 ratios and the distortion component for intermediate ratios. At high ratios the behavior cannot be explained by dominance of either component.

Effects of hearing status and perturbation with a bite block on vowel production

This study explores the effect of hearing status on adaptation to a bite block in vowel productions of normal hearing (NH) adults and adults who use cochlear implants (CI). CI speakers are tested prior to and following experience with the implant. Different sized bite blocks (BB) are used to create unusual degrees of mandibular opening for vowel productions in an /hVd/ context (had, head, heed, hid, and hod). Four conditions are elicited from each NH and CI speaker: (1) no BB with hearing (CI processor on), (2) no BB with no hearing (NH speakers with masking noise and CI speakers with processor off), (3) BB with no hearing, and (4) BB with hearing. Prior to fitting with the implant, CI speakers are tested without hearing in two conditions: (1) no BB and (2) BB. Spectra of the vowel productions are analyzed for dispersion of tokens in the F1–F2 plane in the four conditions. Pilot results support the hypothesis that prior to fitting, CI users are less able to adapt to perturbations than NH speakers and that experience with a CI improves adaptation. The current study is exploring this result further with larger groups of subjects. [Research supported by NIH.]

Digital acoustic analysis of five vowels in maxillectomy patients.

The aim of the study was to characterize the acoustics of vowel articulation in maxillectomy patients. Digital acoustic analysis of five vowels, /a/, /e/, /i/, /o/ and /u/, was performed on 12 male maxillectomy patients and 12 normal male individuals. A simple set of acoustic descriptions called the first and second formant frequencies, F1 and F2, were employed and calculated based on linear predictive coding. The maxillectomy patients had a significantly lower F2 for all five vowels and a significantly higher F1 for only /i/ vowel. From the data plotted on an F1-F2 plane in each subject, we determined the F1 range and the F2 range, which are the differences between the minimum and the maximum frequencies among the five vowels. The maxillectomy patients had a significantly narrower F2 range than the normal controls. In contrast, there was no significant difference in the F1 range. These results suggest that the maxillectomy patients had difficulty in controlling F2 properly. In addition, the speech intelligibility (SI) test was performed to verify the results of this new frequency range method. A high correlation between the F2 range and the score of SI test was demonstrated, suggesting that the F2 range is effective in evaluating the speech ability of maxillectomy patients.

Critical spectral regions for vowel identification.

The first two formant frequencies (F1 and F2) are the cues important for vowel identification. In the categorization of the naturally spoken vowels, however, there are overlaps among the vowels in the F1 and F2 plane. The fundamental frequency (F0), the third formant frequency (F3) and the spectral envelope have been proposed as additional cues. In the present study, to investigate the spectral regions essential for the vowel identification, untrained subjects performed the forced-choice identification task in response to Japanese isolated vowels (/a, o, u, e, i/), in which some spectral regions were deleted. Minimum spectral regions needed for correct vowel identification were the two regions including F1 and F2 (the first and fourth in the quadrisected F1-F2 regions in Bark scale). This was true even when phonetically different vowels had a similar combination of F1 and F2 frequency components. F0 and F3 cues were not necessarily needed. It is concluded that the relative importance in the spectral region is not equivalent, but weighted on the two critical spectral regions. The auditory system may identify the vowels by analyzing the information of the spectral shapes and the formant frequencies (F1 and F2) in these critical spectral regions.

Snail and bivalve populations analysed in relation to physico-chemical quality of lakes in eastern France

The application of multivariate analysis (Factorial Correspondence Analysis) to 4 synthetic descriptors of water mass and sediment gave a statistical distribution, in the F1 F2 plane, for the 18 lakes in our sample. Based on this analysis, 4 categories of lake were defined. The projection of the 19 descriptors relating to the populations of molluscs as supplementary variables on factorial plane and then the affinity analysis between descriptors and system categories has shown that: This analysis also shows that the conditions existing in the deep zone of the lakes and, in consequence, the way in which they operate, determine not only the composition of deep populations but also the composition of the littoral-dwelling populations.

Correlation between production and perception of English vowels

An experiment has been performed to test the hypothesis that a listener refers to his own articulation for perceiving speech. Formant frequencies of eleven English vowels for ten subjects were measured in two different ways: firstly, from the utterances of the vowels in /h/-vowel-/d/ context produced by these subjects and secondly, by presenting a set of synthesised vowel stimuli in /h/-vowel-/d/ context covering the entire F1–F2 plane to the same subjects for identification. Correlation between the production and perception formant frequencies were calculated when the ordering of subjects for the sets of frequencies was the same (i.e., within-subjects correlation) and when it was different (i.e., between-subjects correlation). The difference between the within-subjects correlation and the between-subjects correlation was found not to be statistically significant. Different transformations (e.g. Hertz to Bark, or to mel) and different two-formant models failed to increase significantly the within-subjects correlation with respect to the between-subjects correlation. Thus, the results of this experiment tend to reject the hypothesis.

Nasal vowel identification using LPC‐based formant analysis

An experiment was carried out to determine the problems associated with the identification of nasal vowels in English in the context of vowel recognition using LPC‐based formant analysis. A number of tokens of three pairs of oral/nasal vowels [æ, ɛ, ɪ,] produced in similar phonetic contexts (such as in camp versus camp) were obtained from three male speakers. The first and second formant frequencies obtained as LPC spectral peaks from these pairs of vowels showed significant differences in distribution on the F1−F2 plane. The first formant frequencies of nasal vowels especially had apparent overlap with those of other nasal vowels, while oral vowels had no overlap with each other. The LPC spectra corresponding to the shifted formants and the pole‐zero pairs introduced by the nasal cavity coupling showed substantial individual differences in shape in the first formant region, including different bandwidths and number of peaks. These facts suggest the difficulty of vowel identification in (1) using the same template for both oral and nasal vowels and (2) identifying an unknown vowel even if one knows it is nasalized.

Multidimensional vocalic perceptual space: How many dimensions?

In published sources, the number of recovered dimensions ranges from 2 to 6. From our experiment on 40 French subjects scaling 10 vowels [i,e,ɛ,a,ɔ,o,u,y,ø;,œ] and from a reanalysis of 4 studies [Hanson (1967); Mohr and Wang (1968); Singh and Woods (1971); Terbeeck (1977)] using both metric (Indscal) and nonmetric (Kruskal) methods, the following conclusions can be drawn: (a) Variance accounted for can reach 90% for two‐dimensional group data, leaving little room for more than a third dimension. (b) The first two (unrotated) dimensions and the F1−F2 plane are highly correlated. The most reliable criterion for the presence of a third dimension seems to be the stability of stimuli ordering for repeated analyses from random starting configurations with both methods. Retroflex [vů] and front (nonhigh) rounded [ø,œ] vowels consistently end up at one end of the axis. No ceiling effect of similarities versus distances can explain this ordering for the nonmetric approach. Tentatively, the common acoustical feature between [vů,ø,œ], i.e., the closeness between F2 and F3, is suggested as the third dimension correlate. No evidence for a 4th dimension has been found.

Speech training systems using the lateral shape of the vocal tract and an F1−F2 diagram for hearing‐impaired children

Three speech training systems for hearing‐impaired children were designed and constructed using a minicomputer and a microprocessor. The first system displays the lateral shape of the vocal tract for each vowel estimated from the speech sound. In this system, three storages are prepared. One of them can be used to store a reference shape which may be estimated from a teacher's voice or a prepared standard shape. A child can articulate seeing the reference shape and rearticulate comparing his own first articulation with the reference shape. The second system displays successively the lateral shapes for articulation of any phoneme sequences containing consonants. The lateral shapes for Japanese phoneme sequences are synthesized by a speech synthesizer using an articulatory model. The third system displays the first two formant frequencies extracted from speech as a spot on the F1−F2 plane, where the regions of vowels are also shown. Preliminary use of these devices has shown that they complement each other ...

A developmental study of speech production: Data on vowel imitation and sentence recitation

Ten subjects in each of five groups (4-, 6-, 10-, and 12-year-old children as well as young adults) participated in a study involving the imitation of 15 synthesized vowels and the repetition of three simple sentences. This report emphasizes spectrographic data for the two youngest groups and the adult group. In the vowel imitation task, the subjects could recreate the relative formant structure of the targets, but the children in particular tended to produce allen (non-English) vowels as the proximal English vowels. Plotting of the formant data for different age-sex groups in the F1−F2 plane revealed linear orientations, i.e., isovowel lines for unnormalized data. Standard deviations for the first two formant frequencies of the imitations were commensurate with published data on difference limens for formant frequency. The sentence repetitions were used primarily to evaluate the stability of temporal patterns. Both the inter- and intrasubject standard deviations of temporal segments tended to be larger for the children than for the adults, but even some 4-year-olds occasionally fell within the adult range of variability. [Work supported by NINCDS.]