Anthony Bladon

Production and perception of sibilant fricatives: Shona data

This research began out of two overlapping motives. First, we have been observing what have impressionistically been termed ‘whistling’ fricatives in Shona (and also, though not reported here, in another language, Jibbali); we felt that there was room for a fuller analysis of some aspects of their production characteristics. Second, sibilance in general offers plenty of scope for what Delattre called the ‘quest of the Holy Grail’ in phonetics: the search for features of relevance to perception. More background to both the production and the perception issues will be given as we proceed.

Diphthongs: A case study of dynamic auditory processing

Abstract The dynamic processing of speech in the auditory system is apparently performed in parallel to spectral shape analysis and detection of spectral change. This formulation, closely paraphrased on Chistovich et al. (1982), is a background assumption against which we examine the auditory role of diphthongs as an example of spectral change over time. When we do this, we find a clear conflict in the literature concerning the relative perceptual weight of different components of a diphthong. Does the listener attend to a diphthongs endpoints, or to its spectral rate-of-change? This paper undertakes further experiments to try to resolve the conflict on this issue. As a result of our findings, a simple cumulative model of auditory processing of the speech signal turns out to be inadequate. So too do those models which rely on the rate-of-change itself as a trigger for the spectral comparison process. Instead, the role of spectral change in diphthongs seems to be (a) as a flag commanding extra perceptual weighting by virtue of the fact that change (of some kind) occurs, and (b) as a pointer to adjacent temporal regions of the signal which are important and should be sampled more densely.

Attempts by human speakers to reproduce Fant's nomograms

Abstract Fants nomograms, which constitute a simulation of the acoustic resonances which result from different vocal tract shapes, were checked against natural vowels by two phoneticians carefully producing a full range of vowel articulation in which (a) the location of the minimum aperture in the vocal tract and (b) the lip aperture were separately varied. Spectrographic measurements of formant frequencies were made, some with synchronised lip photographs. Ultrasonic and radiographic methods were used to verify vocal tract constriction size. In many major respects our data agreed with Fants predictions, but three main kinds of discrepancies were noted: firstly, we could not produce as great a range of variations as occurs in the nomograms; secondly, when an [i]-like constriction is moved as far forward as possible, the frequency of F2 does not decrease; and thirdly, variations in lip rounding affect high front vowels and high back vowels differently with no gradual change from one class to the other as in Fants data.

One‐stage and two‐stage temporal patterns of velar coarticulation

Previous observations of the coarticulatory spread of nasality from a nasal consonant to adjacent vowels have yielded some contradictory evidence and two contradictory hypotheses: (a) coarticulatory timing is determined by segmental content and (hi coarticulation is coordinated in time with a particular articulatory gesture. Using transillumination of the velopharyngeal port, the timing patterns of velum opening were investigated in (C)VN, (C)VVN, and (C)VVVN sequences in languages lacking contrastively nasalized vowels. It was found that all speakers used two distinct patterns of velum opening, though with different frequencies of occurrence: first, a single opening gesture of smoothly increasing amplitude, its onset aligned with the first vowel onset irrespective of intervening boundaries; and second, a two‐stage opening gesture whose absolute onset was aligned as previously but whose higher‐velocity second stage was coordinated with the oral closing gesture for the nasal consonant. Thus we may have to ...

Developing computerized transcription exercises for American English

Acquiring the skill of phonetic transcription from orthographic text is a widespread technique included in all undergraduate curricula for phonetics, linguistics and speech science. It is an ability which is also frequently expected in students of modern languages and English. Any instructor who has been faced with a pile of thirty and more transcriptions to mark knows what a laborious task this is, requiring a lot of close examination of fine detail. To be able to lighten the load by computerizing some of the detail of error-correcting is therefore seen as desirable from several points of view.

Spectral edge orientation as a discriminator of fricatives

Both “physical” and “auditorily transformed” spectra of English sibilant fricatives were investigated for evidence that might: (a) discriminate between /s/ and /∫/; (b) do so independently of vowel context; and (c) normalize the substantial speaker‐sex differences found. Measurements included the frequency location of the first peak of fricative noise, the location of the low‐frequency edge descending from this peak, the amplitude range of this descent, and its gradient or orientation. Analyses of variance on our data demonstrated a marked superiority for the orientation of the spectral edge, which differentiated the phonemes reliability, while remaining independent of vowel context or speaker sex. This was true. however, only of the edge as transformed in auditory space. Given the analogous ability of the eye‐to‐track edges, the notation appears to have some plausibility for speech perception theory and an obvious implication for automatic speech recognition. Perceptual confirmation of the importance of ...

A further test of a two‐formant model

The concept of F2′ as an upward weighting of F2 to allow for the perceptual contribution of upper formants [F2′ being calculated according to the formula proposed by R. A. W. Bladon and G. Fant (Quarterly Progress and Status Report, R. I. T. Stockholm, 1978–1, 1–8)] is reviewed critically in the light of the demonstration that vowels exist which sound different although they have the same F1 and F2′. A procedure for generating vocal tract shapes from such formant frequencies yielded shapes which correspond to humanly realizable vowels and yet which are markedly different shapes. Three pairs of vowels synthesized with these formant frequencies were judged by listeners to be auditorily as different as the pair [i‐e] which is phonemically contrastive in many languages. It is concluded that this two‐dimensional model of vowel perception is inadequate, at least as currently formulated.

Synthesis of names using careful speech style

A high‐quality formant synthesizer, optimized for the transfer characteristics of the telephone network, has been implemented to provide real‐time text‐to‐speech performance on a fraction of a TMS 320C25 processor. A version optimized for the pronunciation of American names (‘‘Name‐to‐Speech’’) has been created which incorporates phonetic level changes characteristic of a careful speech style. Examples of the careful speech adaptations include a brief inter‐word silence, a wider use of vowel‐onset glottal stop, the restoration of plosive releases and of various elisions, the strengthening of some weak vowels, stress changes, reductions in formant coarticulation, and various segmental acoustic adjustments such as fricative gain. These careful speech features were determined after observing corresponding behaviors in an initial study of the way telephone users speak their own names (for the purpose of annotating their voice mail). In a follow‐up study, users showed an overwhelming preference for the pronunc...

Extending the Search for a Psychophysical Basis for Dynamic Phonetic Patterns

A prominent flavour of the research arguments summarized in several papers at this Workshop (Diehl, Howell and Rosen, Massaro, Pastore, and discussion sessions) has in my opinion been the following: (1) examine certain well-known phenomena of speech perception, (2) attempt to show that effects of an analogous kind can be demonstrated outside the human speech realm, then (3) argue that the phenomena in question owe their basis not to any ‘specialness’ of the speech code but to properties of auditory analysis more generally. This treatment, as the current volume reveals, has been meted out to such phenomena as categorical perception, fricative versus affricate perception, and voice onset time distinctions. In this way, a limited class of dynamic behaviours in speech, all of them with a long pedigree of experimental investigation (albeit predominantly in English), have been shown to be compatible with auditory-perceptual capabilities.

Analysis and perception of sibilant fricatives: Shona data

In addition to the commonly found [s,z] and [∫ȝ], the Shona language has a third pair of sibilants [s,zw w] usually described as labialized alveolar fricatives and sometimes called “whistling” fricatives. Analysis of all these sibilants revealed that the location of the main frequency peak was typically 3 Bark higher for [s,z] than for the remaining four fricatives which did not differ significantly in this respect. Durational differences were slight throughout, and seemed unlikely to contribute as place cues. A likelier possibility for cueing the place distinction, we hypothesized, was an interaction of peak location with the gradient of the high‐ and low‐frequency slopes adjacent to the peak, together perhaps with contributions from the VCV transitions. Experimental manipulation of these variables in synthetic versions of the three voiceless sibilants yielded listener identifications suggesting that (1) a flat HF slope cued [s] overridingly, irrespective of the peak location, but (2) peak location becam...