Cues For Place Of Articulation Research Articles

Formant transitions as acoustic cues for place of articulation in speech perception

Formant transitions as acoustic cues for place of articulation in speech perception

Perception of voiceless nasals in Mizo and Angami

This study is a perceptual analysis of place of articulation of voiceless nasals and their voiced counterparts in Mizo and Angami, two Tibeto-Burman languages spoken in North-East India. The voiceless nasals in Mizo and Angami differ in terms of their characteristics of voicing. In Mizo, there is a portion of voiced nasal murmur before the vowel starts, which has been compared to the voiceless nasals in Burmese (Bhaskararao & Ladefoged,1991). However, in Angami, the nasal murmur remains voiceless throughout (Bhaskararao & Ladefoged,1991; Blankenship et al.,1993). A previously performed acoustic analysis of cues for place of articulation in both the languages shows that the cues were more robust in the transition portion of the vowel following the voiceless nasal (Gogoi, 2018). The perception of voiceless nasals by native speakers have not yet been studied. However, existing literature on Burmese voiceless nasals have shown that there are sufficient acoustic cues in the voiced murmur portion of the voiceless nasals for discrimination of the place of articulation (Dantsuji,1986). This study investigates whether the perception results correspond to the acoustic analysis previously observed.

Inter- and intra-speaker variability in the production of voiceless nasal consonants

This study uses both acoustic and aerodynamic data to investigate variation in the realization of phonemically voiceless nasal consonants, which have been found to differ cross-linguistically. In particular, it examines the timing of voicing and oral and nasal air flow during production of voiceless nasal consonants in both word-initial and word-medial positions in Hakha Chin, a Tibeto-Burman language spoken in western Burma/Myanmar. Comparing three other Tibeto-Burman languages, Bhaskararao and Ladefoged (1991) found two distinct patterns in the production of voiceless nasal consonants, which varied by language, but reported general consistency among speakers of the same language. Results of the present study show considerable variation in the realization of voiceless nasals, both between individual speakers and even between individual tokens produced by the same speaker. This high degree of variability may be linked to the relatively weak acoustic cues for place of articulation among nasal consonants in general and for voiceless nasals in particular.This study uses both acoustic and aerodynamic data to investigate variation in the realization of phonemically voiceless nasal consonants, which have been found to differ cross-linguistically. In particular, it examines the timing of voicing and oral and nasal air flow during production of voiceless nasal consonants in both word-initial and word-medial positions in Hakha Chin, a Tibeto-Burman language spoken in western Burma/Myanmar. Comparing three other Tibeto-Burman languages, Bhaskararao and Ladefoged (1991) found two distinct patterns in the production of voiceless nasal consonants, which varied by language, but reported general consistency among speakers of the same language. Results of the present study show considerable variation in the realization of voiceless nasals, both between individual speakers and even between individual tokens produced by the same speaker. This high degree of variability may be linked to the relatively weak acoustic cues for place of articulation among nasal consonants in...

The effect of viewing speech on auditory speech processing is different in the left and right hemispheres

We used whole-head magnetoencephalograpy (MEG) to record changes in neuromagnetic N100m responses generated in the left and right auditory cortex as a function of the match between visual and auditory speech signals. Stimuli were auditory-only (AO) and auditory–visual (AV) presentations of /pi/, /ti/ and /vi/. Three types of intensity matched auditory stimuli were used: intact speech (Normal), frequency band filtered speech (Band) and speech-shaped white noise (Noise). The behavioural task was to detect the /vi/ syllables which comprised 12% of stimuli. N100m responses were measured to averaged /pi/ and /ti/ stimuli. Behavioural data showed that identification of the stimuli was faster and more accurate for Normal than for Band stimuli, and for Band than for Noise stimuli. Reaction times were faster for AV than AO stimuli. MEG data showed that in the left hemisphere, N100m to both AO and AV stimuli was largest for the Normal, smaller for Band and smallest for Noise stimuli. In the right hemisphere, Normal and Band AO stimuli elicited N100m responses of quite similar amplitudes, but N100m amplitude to Noise was about half of that. There was a reduction in N100m for the AV compared to the AO conditions. The size of this reduction for each stimulus type was same in the left hemisphere but graded in the right (being largest to the Normal, smaller to the Band and smallest to the Noise stimuli). The N100m decrease for the Normal stimuli was significantly larger in the right than in the left hemisphere. We suggest that the effect of processing visual speech seen in the right hemisphere likely reflects suppression of the auditory response based on AV cues for place of articulation.

A study of regressive place assimilation in spontaneous speech and its implications for spoken word recognition

Regressive place assimilation is a form of pronunciation variation in which a word-final alveolar sound takes the place of articulation of a following labial or velar sound, as when green boat is pronounced greem boat. How listeners recover the intended word (e.g., green, given greem) has been a major focus of spoken word recognition theories. However, the extent to which this variation occurs in casual, unscripted speech has previously not been reported. Two studies of pronunciation variation were conducted using a spontaneous speech corpus. First, phonetic labeling data were used to identify contexts in which assimilation could occur, namely, when a word-final alveolar stop (/t/, /d/, or /n/) was followed by a velar or labial consonant. Assimilation was indicated relatively infrequently, while deletion, glottalization, or canonical pronunciations were more often indicated. Moreover, lexical frequency was shown to affect pronunciation; high frequency lexical items showed more types of variation. Second, acoustic analyses showed that neither place of articulation cues (indicated by second formant variation) nor relative amplitude was sufficient to distinguish assimilated from deleted and canonical variants; only when closure duration was additionally taken into account were these three variant types distinguishable. Implications for theories of word recognition are discussed.

Perceptual Weighting of Stop Consonant Cues by Normal and Impaired Listeners in Reverberation Versus Noise

To determine if listeners with normal hearing and listeners with sensorineural hearing loss give different perceptual weightings to cues for stop consonant place of articulation in noise versus reverberation listening conditions. Nine listeners with normal hearing (23-28 years of age) and 10 listeners with sensorineural hearing loss (31-79 years of age, median 66 years) participated. The listeners were asked to label the consonantal portion of synthetic CV stimuli as either /p/ or /t/. Two cues were varied: (a) the amplitude of the spectral peak in the F4/F5 frequency region of the burst was varied across a 30-dB range relative to the adjacent vowel peak amplitude in the same frequency region, (b) F2/F3 formant transition onset frequencies were either appropriate for /p/, /t/ or neutral for the labial/alveolar contrast. Weightings of relative amplitude and transition cues for voiceless stop consonants depended on the listening condition (quiet, noise, or reverberation), hearing loss, and age of listener. The effects of age with hearing loss reduced the perceptual integration of cues, particularly in reverberation. The effects of hearing loss reduced the effectiveness of both cues, notably relative amplitude in reverberation. Reverberation and noise conditions have different perceptual effects. Hearing loss and age may have different, separable effects.

Frequency band-importance functions for auditory and auditory-visual speech recognition

In many everyday listening environments, speech communication involves the integration of both acoustic and visual speech cues. This is especially true in noisy and reverberant environments where the speech signal is highly degraded, or when the listener has a hearing impairment. Understanding the mechanisms involved in auditory-visual integration is a primary interest of this work. Of particular interest is whether listeners are able to allocate their attention to various frequency regions of the speech signal differently under auditory-visual conditions and auditory-alone conditions. For auditory speech recognition, the most important frequency regions tend to be around 1500–3000 Hz, corresponding roughly to important acoustic cues for place of articulation. The purpose of this study is to determine the most important frequency region under auditory-visual speech conditions. Frequency band-importance functions for auditory and auditory-visual conditions were obtained by having subjects identify speech tokens under conditions where the speech-to-noise ratio of different parts of the speech spectrum is independently and randomly varied on every trial. Point biserial correlations were computed for each separate spectral region and the normalized correlations are interpreted as weights indicating the importance of each region. Relations among frequency-importance functions for auditory and auditory-visual conditions will be discussed.

On the effectiveness of whole spectral shape for vowel perception.

The formant hypothesis of vowel perception, where the lowest two or three formant frequencies are essential cues for vowel quality perception, is widely accepted. There has, however, been some controversy suggesting that formant frequencies are not sufficient and that the whole spectral shape is necessary for perception. Three psychophysical experiments were performed to study this question. In the first experiment, the first or second formant peak of stimuli was suppressed as much as possible while still maintaining the original spectral shape. The responses to these stimuli were not radically different from the ones for the unsuppressed control. In the second experiment, F2-suppressed stimuli, whose amplitude ratios of high- to low-frequency components were systemically changed, were used. The results indicate that the ratio changes can affect perceived vowel quality, especially its place of articulation. In the third experiment, the full-formant stimuli, whose amplitude ratios were changed from the original and whose F2's were kept constant, were used. The results suggest that the amplitude ratio is equal to or more effective than F2 as a cue for place of articulation. We conclude that formant frequencies are not exclusive cues and that the whole spectral shape can be crucial for vowel perception.

Place of articulation and first formant transition pattern both affect perception of voicing in English

Voice onset time (VOT) and first-formant (F1) transition characteristics are both important phonetic cues for voicing categorization of pretonic stops in English. These cues also covary with place of articulation in production, such that bilabials are produced with shorter VOT values, and velars are produced with longer VOT values. This study investigated the effects of place of articulation and F1transition pattern on voicing categorization of VOT continua. Listeners categorized for voicing and place of articulation the stop consonant in VCV and CV stimuli for which the variables F1transition pattern, place of articulation (bilabial, alveolar, and velar), and VOT were fully crossed. A logistic regression analysis shows that bilabial and alveolar stimuli were more likely than velar stimuli to be classified as voiceless, largely independent of F1transition pattern. Increasing F1onset frequency and shortening transition duration also made voiceless judgments more likely. The magnitude of theF1 transition effects was considerably larger than the place of articulation effect. The results are inconsistent with models of consonant classification in which acoustic–phonetic cues for place of articulation are not involved in the perception of the voicing contrast. However, the observed perceptual interaction between place of articulation and voicing may be consistent with either a feature- or segment-based model of consonant classification.

Cues for place and palatalization in Russian stops

Using acoustic measurements and native speaker perceptions, the cues for place of articulation of Russian plain and palatalized stops were investigated. The acoustic measurements were made of formant transitions, burst duration, and spectral burst shape. They showed that formant transitions distinguish place of articulation of plain consonants, but not palatalized consonants. Differences distinguishing place of articulation were found in the burst durations and in the ratio of energy between the upper and lower portions of the burst spectra. In the perception experiment, native Russian speakers listened to various consonant–vowel stimuli. Hearing only the formant transitions, listeners were able to identify place of articulation of just the plain stops. When they heard stimuli with conflicting cues for place in the burst and formant transitions, in the majority of the cases, they identified plain consonants based on the formant transitions, but they identified palatalized consonants based on the burst. It is concluded that while formant transitions do provide the strongest cues for place among plain stops, they do not provide cues for place among palatalized stops. The cues for place of articulation of palatalized consonants are contained completely in the duration and spectral shape of the burst.

Evaluation of simulations of the effects of hearing loss produced by combinations of dynamic expansion and spectral smearing

By processing sounds with combinations of multiband expansion (MBE) and spectral smearing (SS), listeners with normal hearing can experience the elevated thresholds, abnormal growth of loudness, and reduced frequency selectivity similar to those of listeners with sensorineural hearing impairments. The effects of simulations of flat 40-dB losses produced by such combinations on masking patterns and loudness summation, and of simulations of a more severe loss on speech perception in noise, were studied. Both MBE and SS reduce frequency selectivity as measured by masking patterns for tones produced by narrow-band noise, and both broaden forward-masked tuning curves for tones. However, for equally loud probe tones, MBE elevates the tips of psychoacoustic tuning curves much more than SS. As measured by loudness balances between narrow-band and wideband noises, MBE reduces loudness summation much more than SS. The effects of MBE and SS simulations of a severe hearing loss on consonant identification in speech-shaped noise depend on the SNR, with MBE more detrimental at high SNR and SS more detrimental at low SNR. MBE tends to degrade reception of cues for voicing; SS, cues for place of articulation; both degrade cues for manner of articulation. [Work supported by NIH.]

Transition duration effects on place perception in hearing-impaired listeners.

The purpose of this study was to determine the contribution of transition duration in signaling an alveolar/velar place-of-articulation contrast to listeners with mild-to-moderate sensorineural hearing loss. In the experiments reported here, F1, F2, and F3 transitions were varied between values appropriate for [deh] and [geh]. These were combined with transition durations ranging from 20 to 50 ms in 5-ms steps. In three separate identification tasks, listeners were presented with: (1) all levels of formant pattern varied orthogonally with all transition duration values; (2) all levels of formant pattern combined with an ambiguous duration value; and (3) all levels of transition duration combined with an ambiguous formant pattern. A trend was observed for hearing-impaired listeners to identify stimuli with shorter transition durations as [deh] in the absence of other unequivocal cues for place of articulation. Overall performance of hearing-impaired listeners was more variable than that of normal listeners, predicted in part by the results of Summerfield etal. [Speech Commun. 4, 213–229 (1985)], in which hearing-impaired listeners showed a decreased ability to identify place of articulation from transition information alone. Results of these experiments will be discussed in terms of current views of temporal processing in hearing-impaired listeners. [Work supported by NIH.]

Are F2‐ and F3‐onset frequencies equal cues for place of articulation

This study reevaluated F2‐ and F3‐onset frequencies as cues for place of articulation. Seven levels of F2‐onset frequency, ranging from 600 to 1800 Hz, were factorially combined with eight levels of F3‐onset frequency, ranging from 1400 to 4200 Hz, to create 56 synthetic syllables. Subjects were asked to classify each initial phoneme of the syllable into/ba/, /da/,/ga/, or/others/. A canonical correlation analysis was conducted to recover the relationship between the percentage of each response category and the values of F2‐ and F3‐onset frequencies. The canonical correlation structure shows that F2‐onset frequency has a high negative correlation with the /ba/ category, and a positive correlation with the/da/ category, while both F2‐ and F3‐onset frequencies are moderately correlated with the /ga/ category. This result suggests that F2‐onset frequency is an adequate cue to distinguish /ba/ from /da/. However, both F2‐ and F3‐onset frequencies are interacted to cue /da/ and /ga/ distinction. This result is in contrast to claims that either F2‐ or F3‐onset frequency is a strong cue for three place categories.

Speech changes in a postlingually deafened girl using a multichannel cochlear implant

The speech of a postlingually deafened girl previously implanted with a single electrode implant was spectrographically analyzed from recordings taken on six separate occasions; on three occasions after her single electrode implant failed (4 days post, 5 days post, 18 days post) and on three occasions after stimulation of the Nucleus multichannel implant (1 day post, 6 months post, 9 months post). Changes in acoustic measurements of vowel formants, vowel duration, voicing, manner of articulation, and place of articulation cues are discussed. Listener judgments of target words are presented. The present findings are related to previous case studies of speech changes with a multichannel cochlear implant [Tartter et al., J. Acoust. Soc. Am. 86, 2113–2121 (1989)] and single electrode stimulation [Leder et al., J. Acoust. Soc. Am. 79, 1967–1974 (1986)].

Perceptual cues for place of articulation for the voiced pharyngeal and uvular consonants

A series of perceptual experiments was conducted to investigate the perceptual cues for place of articulation for the voiced pharyngeal /// and the voiced uvular /3/ consonants in Arabic. Synthetic consonant–vowel stimuli, in which the vowel was the long vowel /aa/ of Arabic, were used in the experiments. In the first experiment, stimulus continua were constructed by systematically varying the trajectories of the first and second formant frequencies (F1 and F2) independently between the values appropriate for the two consonants. In the second experiment, the bandwidth of F2 for the pharyngeal stimuli and the bandwidth of F1 for the uvular stimuli were manipulated. Results of the first experiment show that the onset value of F1(F1o) is essential in discriminating between the two consonants, while that of F2 is not. Results of the second experiment show that widening the bandwidth of F1 increases the percent of uvular responses and enhances the naturalness of the uvular stimuli, whereas it decreases substantially the number of pharyngeal responses. Increasing the bandwidth of F2 had no effect on either the percent identification of the pharyngeal or the naturalness of the synthetic stimuli.

Invariant acoustic cues in stop consonants: A cross‐language study using the Wignet distribution

Acoustic invariance was investigated in 200 unaspirated stops from English, Telugu, and French, using the Wigner distribution. The bilabials were characterized either by absence of any noticeable burst, or by an energy concentration in the low‐frequency region during the burst, when present. The burst of alveolars was mostly diffuse in nature, but sometimes became compact, especially when it was followed by a back vowel. The burst of velars had a compact spectrum centered near F2 of the vowel, when followed by back vowels, but was most often diffuse when followed by a front vowel. The burst of all dental from Telugu was found to be diffuse. When the burst of alveolars was compact, or that of velars diffuse, the duration from the start of the burst to the start of the vowel and F2 of the vowel were found to be reliable cues for place of articulation. The diffuse nature of bilabials and alveolars, reported by Lahiri et al. [J. Acoust. Soc. Am. 76, 391–404 (1984)] and by the authors referenced therein, can be partly attributed to the temporal averaging of the burst spectrum and the diffuse vowel spectrum, which is inevitable in Fourier‐type analyses [Garudadri et al., J. Acoust. Soc. Am. Suppl. 1 79, S94 (1986)]. Unlike in other investigations, no apparent pattern could be found for the tokens wrongly classified. [Work supported, in part, by NSERC, Canada.]

Effects of voicing, place, and vowel context on VOT for French and English stops

Preliminary results in a study of VOT in the production of French initial stops (recorded from eight native speakers) indicate a significant interaction of voicing state, place of articulation, and vowel identity in /CVk/syllables. While some of the trends observed are similar (e.g., VOT for high vowels is generally longer for that of non‐high vowels), the exact pattern of interactions does not agree with that found for American English (R. F. Port and R. Rutonno, J. Acoust. Soc. Am. 66, 654–662 (1979)]. Results from a larger‐scale measurement study comparing VOT production in French and Canadian English will be discussed. Interaction effects that depend explicitly on categorical (as opposed to purely acoustic) context have been shown to be reliable in VOT production experiments [Port and Rutonno (1979)]. Results of cross‐language perceptual experiments (using VOT continua in which vocalic context and place of articulation cues are varied as parameters) designed to test the perceptual reliablity of such effects will be discussed. [Work supported by SSHRC.]

Speech perception and language acquisition

Voice onset time (VOT) and first-formant (F1) transition characteristics are both important phonetic cues for voicing categorization of pretonic stops in English. These cues also covary with place of articulation in production, such that bilabials are produced with shorter VOT values, and velars are produced with longer VOT values. This study investigated the effects of place of articulation and F1transition pattern on voicing categorization of VOT continua. Listeners categorized for voicing and place of articulation the stop consonant in VCV and CV stimuli for which the variables F1transition pattern, place of articulation (bilabial, alveolar, and velar), and VOT were fully crossed. A logistic regression analysis shows that bilabial and alveolar stimuli were more likely than velar stimuli to be classified as voiceless, largely independent of F1transition pattern. Increasing F1onset frequency and shortening transition duration also made voiceless judgments more likely. The magnitude of theF1 transition effects was considerably larger than the place of articulation effect. The results are inconsistent with models of consonant classification in which acoustic–phonetic cues for place of articulation are not involved in the perception of the voicing contrast. However, the observed perceptual interaction between place of articulation and voicing may be consistent with either a feature- or segment-based model of consonant classification.

Discrete representation of perceptual structure underlying consonant confusions.

The perceptual representation of speech is generally assumed to be discrete rather than continuous, pointing to the need for general discrete analytic models to represent observed perceptual similarities among speech sounds. The INDCLUS (INdividual Differences CLUStering) model and algorithm [J.D. Carroll and P. Arabie, Psychometrika 48, 157-169 (1983)] can provide this generality, representing symmetric three-way similarity data (stimuli X stimuli X conditions) as an additive combination of overlapping, and generally not hierarchial, clusters whose weights (which are numerical values gauging the importance of the clusters) vary both as a function of the cluster and condition being considered. INDCLUS was used to obtain a discrete representation of underlying perceptual structure in the Miller and Nicely consonant confusion data [G.A. Miller and P.E. Nicely, J. Acoust. Soc. Am. 27, 338-352 (1955)]. A 14-cluster solution accounted for 82.9% of total variance across the 17 listening conditions. The cluster composition and the variations in cluster weights as a function of stimulus degradation were interpreted in terms of the common and unique perceptual attributes of the consonants within each cluster. Low-pass filtering and noise masking selectively degraded unique attributes, especially the cues for place of articulation, while high-pass filtering degraded both unique and common attributes. The clustering results revealed that perceptual similarities among consonants are accurately modeled by additive combinations of their specific and discrete acoustic attributes whose weights are determined by the nature of the stimulus degradation.

Perceptual integration of the murmur and formant transitions for place of articulation in nasal consonants.

This study reassessed the role of the nasal murmur and formant transitions as perceptual cues for place of articulation in nasal consonants across a number of vowel environments. Five types of computer-edited stimuli were generated from natural utterances consisting of [m n] followed by [i e a o u]: (1) full murmurs; (2) transitions plus vowel segments; (3) the last six pulses of the murmur; (4) the six pulses starting from the beginning of the formant transitions; and (5) the six pulses surrounding the nasal release (three pulses before and three pulses after). Results showed that the murmur provided as much information for the perception of place of articulation as did the transitions. Moreover, the highest performance scores for place of articulation were obtained in the six-pulse condition containing both murmur and transition information. The data support the view that it is the combination of nasal murmur plus formant transitions which forms an integrated property for the perception of place of articulation.