Showing papers on "Voice published in 2008"

Positional faithfulness and voicing assimilation in optimality theory

Abstract: This paper proposes a set of constraints within the framework of Optimality Theory that accounts for syllable-final laryngeal neutralization and voicing assimilation in obstruent clusters. The interaction of positional faithfulness and markedness is shown to result in laryngeal neutralization. Regressive assimilation is shown to be a result of the interaction of positional faithfulness with a constraint preferring adjacent obstruents to agree in voicing. Rerankings of the proposed constraints account for attested patterns of voicing assimilation and neutralization: unlike previous neutralization and spread analyses, this approach makes the correct prediction that it is equally natural for voicing assimilation in clusters to combine with either devoicing of or retention of voicing distinctions in word-final consonants. It is argued that the interactions of these constraints account for why voicing assimilation is always regressive unless special circumstances hold.

Cue-specific effects of categorization training on the relative weighting of acoustic cues to consonant voicing in English

Abstract: In English, voiced and voiceless syllable-initial stop consonants differ in both fundamental frequency at the onset of voicing (onset F0) and voice onset time (VOT). Although both correlates, alone, can cue the voicing contrast, listeners weight VOT more heavily when both are available. Such differential weighting may arise from differences in the perceptual distance between voicing categories along the VOT versus onset F0 dimensions, or it may arise from a bias to pay more attention to VOT than to onset F0. The present experiment examines listeners’ use of these two cues when classifying stimuli in which perceptual distance was artificially equated along the two dimensions. Listeners were also trained to categorize stimuli based on one cue at the expense of another. Equating perceptual distance eliminated the expected bias toward VOT before training, but successfully learning to base decisions more on VOT and less on onset F0 was easier than vice versa. Perceptual distance along both dimensions increased for both groups after training, but only VOT-trained listeners showed a decrease in Garner interference. Results lend qualified support to an attentional model of phonetic learning in which learning involves strategic redeployment of selective attention across integral acoustic cues.

Tracking the time course of phonetic cue integration during spoken word recognition.

Abstract: Speech perception requires listeners to integrate multiple cues that each contribute to judgments about a phonetic category. Classic studies of trading relations assessed the weights attached to each cue but did not explore the time course of cue integration. Here, we provide the first direct evidence that asynchronous cues to voicing (/b/ vs. /p/) and manner (/b/ vs. /w/) contrasts become available to the listener at different times during spoken word recognition. Using the visual world paradigm, we show that the probability of eye movements to pictures of target and of competitor objects diverge at different points in time after the onset of the target word. These points of divergence correspond to the availability of early (voice onset time or formant transition slope) and late (vowel length) cues to voicing and manner contrasts. These results support a model of cue integration in which phonetic cues are used for lexical access as soon as they are available.

Adaptive voice playout in VOP

Abstract: Packetized CELP-encoded speech playout with frame truncation during silence and frame expansion method dependent upon voicing classification with voiced frame expansion maintaining phasealignment.

Phonetic naturalness and unnaturalness in Japanese loanword phonology

Abstract: This paper argues that phonetic naturalness and unnaturalness can interact within a single grammatical system. In Japanese loanword phonology, only voiced geminates, but not voiced singletons, devoice to dissimilate from another voiced obstruent. The neutralizability difference follows from a ranking which Japanese speakers created on perceptual grounds: Ident(voi)Sing » Ident(voi)Gem. On the other hand, the trigger of devoicing—OCP(voi)—has no phonetic underpinning because voicing does not have phonetic characteristics that would naturally lead to confusion-based dissimilation (Ohala, Proceedings of Chicago Linguistic Society: Papers from the parasession on language and behaviour, 1981, in: Jones (ed.) Historical linguistics: Problems and perspectives, 1993). OCP(voi) in Modern Japanese originated as a phonetically natural OCP(prenasal) in Old Japanese because the spread out heavy nasalization would lead to perceptual confusion, but it divorced from its phonetic origin when prenasalization became voicing. The interaction of the three constraints in Modern Japanese suggests that phonetic naturalness (the ranking Ident(voi)Sing » Ident(voi)Gem) and unnaturalness (OCP(voi)) co-reside within a single module.

Production and Perception of Temporal Patterns in Native and Non-Native Speech

Abstract: Two experiments examined production and perception of English temporal patterns by native and non-native participants. Experiment 1 indicated that native and non-native (L1 = Chinese) talkers differed significantly in their production of one English duration pattern (i.e., vowel lengthening before voiced versus voiceless consonants) but not another (i.e., tense versus lax vowels). Experiment 2 tested native and non-native listener identification of words that differed in voicing of the final consonant by the native and non-native talkers whose productions were substantially different in experiment 1. Results indicated that differences in native and non-native intelligibility may be partially explained by temporal pattern differences in vowel duration although other cues such as presence of stop releases and burst duration may also contribute. Additionally, speech intelligibility depends on shared phonetic knowledge between talkers and listeners rather than only on accuracy relative to idealized production norms.

Duration differences in the articulation and acoustics of Swiss German word-initial geminate and singleton stops

Abstract: Stops in Swiss German contrast only in quantity in all word positions; aspiration and voicing play no role. As in most languages with consonant quantity contrast, geminate stops are produced with significantly longer closure duration (CD) than singletons in an intersonorant context. This holds word medially as well as phrase medially, e.g., [oni tto:sə] “without roar” versus [oni to:sə] “without can.” Since the stops are voiceless, no CD cue distinguishes geminates from singletons phrase initially. Nevertheless, do speakers utilize articulatory means to maintain the contrast? By using electropalatography, the articulatory and acoustic properties of word-initial alveolar stops were investigated in phrase-initial and phrase-medial contexts. The results are threefold. First, as expected, CD and contact duration of the articulators mirror each other within a phrase: Geminates are longer than singletons. Second, phrase initially, the contact data unequivocally establish a quantity distinction. This means that—...

Predicting word-recognition performance in noise by young listeners with normal hearing using acoustic, phonetic, and lexical variables.

Abstract: Purpose: To analyze the 50% correct recognition data that were from the Wilson et al (this issue) study and that were obtained from 24 listeners with normal hearing; also to examine whether acoustic, phonetic, or lexical variables can predict recognition performance for monosyllabic words presented in speech-spectrum noise. Research Design: The specific variables are as follows: (a) acoustic variables (i.e., effective root-meansquare sound pressure level, duration), (b) phonetic variables (i.e., consonant features such as manner, place, and voicing for initial and final phonemes; vowel phonemes), and (c) lexical variables (i.e., word frequency, word familiarity, neighborhood density, neighborhood frequency). Data Collection and Analysis: The descriptive, correlational study will examine the influence of acoustic, phonetic, and lexical variables on speech recognition in noise performance. Results: Regression analysis demonstrated that 45% of the variance in the 50% point was accounted for by acoustic and phonetic variables whereas only 3% of the variance was accounted for by lexical variables. These findings suggest that monosyllabic word-recognition-in-noise is more dependent on bottom-up processing than on top-down processing. Conclusions: The results suggest that when speech-in-noise testing is used in a pre- and post-hearingaid-fitting format, the use of monosyllabic words may be sensitive to changes in audibility resulting from amplification.

Et si le proto-bantu était aussi une langue … avec ses contraintes et ses déséquilibres

Abstract: This article aims to define the phonological structure of Proto-Bantu lexical units on the basis of frequencies of reconstructed consonant co-occurrences. Starting from the main reconstructions given in BLR3, I present evidence for the presence of unexpected frequencies indicating imbalances in two directions. Certain consonant co-occurrences have not been reconstructed, essentially consonants sharing the same place of articulation and differing by only one feature, either voicing or nasality. These “gaps” in the proto-lexicon turn out to correspond to more general constraints that tend, on the one hand, towards the differentiation of place of articulation and, on the other hand, on agreement in voicing and nasality. However, in cases where *C1 and *C2 share the same place of articulation, Proto-Bantu seems to prefer identity over similarity. In looking to establish a link between the phonotactic constraints of the mother language and those of daughter languages, the latter take different directions, either a direction identical to that of the mother language, or a divergent one. In the reconstructions, the constraints on the nasality feature show similarities to those present in contemporary languages: Ganda has extended the constraint reconstructed for alveolars to all co-occurrences between a voiced stop and a nasal with the same place of articulation. However, the constraints on voicing generated by the dissimilation rule known as Dahl’s Law go in a divergent direction. I bring support here for the view that Dahl’s Law is in fact a daughter-language innovation. Furthermore, I show that this innovation was probably induced by the imbalances of the mother language; the rule fills Proto-Bantu distributional gaps. Finally, I discuss the implications of this study for the Obligatory Contour Principle (OCP).

Stop consonant voicing and intraoral pressure contours in women and children

Abstract: Previous authors have established that stop consonant voicing is more limited in young children than adults, and have ascribed this to immature vocal-tract pressure management Physical development relevant to speech aerodynamics continues into adolescence, suggesting that consonant voicing development may also persist into the school-age years This study explored the relationship between stop consonant voicing and intraoral pressure contours in women, 5year olds, and 10year olds Productions of intervocalic /p b/ were recorded from eight speakers at each age Measures were made of stop consonant voicing and δ, a measure designed to characterize the time course of intraoral pressure increase in stops, following Muller and Brown [Speech and Language: Advances in Basic Research and Practice, edited by N Lass (Academic, Madison, 1980), Vol 4, pp 318–389] Age effects for stop consonant voicing and δ were not statistically significant, but correlations between δ and stop voicing were less often significan

Information transfer analysis: a first look at estimation bias.

Abstract: Information transfer analysis [G. A. Miller and P. E. Nicely, J. Acoust. Soc. Am. 27, 338–352 (1955)] is a tool used to measure the extent to which speech features are transmitted to a listener, e.g., duration or formant frequencies for vowels; voicing, place and manner of articulation for consonants. An information transfer of 100% occurs when no confusions arise between phonemes belonging to different feature categories, e.g., between voiced and voiceless consonants. Conversely, an information transfer of 0% occurs when performance is purely random. As asserted by Miller and Nicely, the maximum-likelihood estimate for information transfer is biased to overestimate its true value when the number of stimulus presentations is small. This small-sample bias is examined here for three cases: a model of random performance with pseudorandom data, a data set drawn from Miller and Nicely, and reported data from three studies of speech perception by hearing impaired listeners. The amount of overestimation can be substantial, depending on the number of samples, the size of the confusion matrix analyzed, as well as the manner in which data are partitioned therein.

Variation in stop consonant voicing in two regional varieties of American English.

Abstract: This study is an acoustic investigation of the nature and extent of consonant voicing of the stop /b/ in two dialectal varieties of American English spoken in south‐central Wisconsin and western North Carolina. The stop /b/ occurred at the juncture of two words such as small bids, in a position between two voiced sonorants, i.e., the liquid /l/ and a vowel. Twenty women participated, ten representing the Wisconsin and the North Carolina variety, respectively. Significant dialectal differences were found in the voicing patterns. The Wisconsin stop closures were usually not fully voiced and terminated in a complete silence followed by a noisy and voiceless closure release, whereas North Carolina speakers produced mostly fully voiced closures. Further dialectal differences included the proportion of closure voicing as a function of word emphasis. For Wisconsin speakers, the proportion of closure voicing was smallest when the word was emphasized and it was greatest in nonemphatic positions. For North Carolina speakers, the degree of word emphasis did not have an effect on the proportion of closure voicing. The results are discussed in terms of differences in the way voicing is maintained during the closure by the speakers of respective dialects. [Work supported by NIH..]

Preschoolers use partial letter names to select spellings: Evidence from Portuguese

Abstract: Two studies examined children's use of letter–name spelling strategies when target phoneme sequences match letter names with different degrees of precision. We examined Portuguese-speaking preschoolers' use of h (which is named /a′ga/ but which never represents those sounds) when spelling words beginning with /ga/ or variants of /ga/. We also looked at use of q (named /ke/) when spelling /ke/ and /ge/. Children sometimes used h for stimuli beginning with /ga/ and /ka/, and q when spelling words and nonwords beginning with /ke/ and /ge/ they did not use these letters when stimuli began with other sequences. Thus, their spellings evinced use of letter-name matches primarily when consonant-vowel sequences matched, such that vowels must be exact but consonants could differ in voicing from the target phoneme.

Loanword adaptation between Japanese and Korean: evidence for L1 feature-driven perception

Abstract: The purpose of the present study is two-fold. First is to propose that an L1 (host language) speakers’ perception of L2 (donor language) sounds is conditioned by the acoustic cues to the laryngeal features of the L1 grammar in the loanword adaptation of Korean voiceless stops into Japanese and of the Japanese voicing contrast and voiceless geminates into Korean. Second is to suggest the enhancing role of some L2 or L1 phonetic properties in perceiving L2 variant(s) as distinctive according to the system of L1 features in loanword adaptation.

Predicting F0 and voicing from NAM-captured whispered speech

Abstract: The NAM-to-speech conversion proposed by Toda and colleagues which converts Non-Audible Murmur (NAM) to audible speech by statistical mapping trained using aligned corpora is a very promising technique, but its performance is still insufficient, mainly due to the difficulty in estimating F0 of the transformed voice from unvoiced speech. In this paper, we propose a method to improve F0 estimation and voicing decision in a NAM-to-speech conversion system based on Gaussian Mixture Models (GMM) applied to whispered speech. Instead of combining voicing decision and F0 estimation in a single GMM, a simple feed-forward neural network is used to detect voiced segments in the whisper while a GMM estimates a continuous melodic contour based on training voiced segments. The error rate for the voiced/unvoiced decision of the network is 6.8% compared to 9.2% with the original system. Our proposal benefits also to F0 estimation error.

Breathy and whispery voicing in White Hmong

Abstract: The White dialect of Hmong uses breathy voice as a tonal feature, and also a distinctive whispery voice as a stop consonant feature. In this paper, acoustic measurements are shown to validate the apparent differences between these two similar phonation types. In particular, relative harmonic intensity and harmonicity were found to be, in general, three ways distinct among Hmong modal, breathy, and whispery phonation. The discovery of distinctly pronounced breathy and whispery phonation in a single language has implications for the representational theory which is used to specify the phonetic grammar.

Frication and Voicing Classification

Abstract: Phonetic detail of voiced and unvoiced fricatives was examined using speech analysis tools Outputs of eight f0 trackers were combined to give reliable voicing and f0 values Log - energy and Mel frequency cepstral features were used to train a Gaussian classifier that objectively labeled speech frames for frication Duration statistics were derived from the voicing and frication labels for distinguishing between unvoiced and voiced fricatives in British English and European Portuguese

Features for automatic detection of voice bars in continuous speech.

Abstract: In this paper we propose features for automatic detection of voice bar, which is an essential component of voiced stop consonants, in continuous speech. The acoustic-phonetic and production based knowledge such as, the presence of voicing, low strength of excitation compared to other voiced phones and a predominant low-frequency spectral energy, are mapped onto a set of acoustic features that can be automatically extracted from the signal. The usefulness of the proposed features in the detection of voice bars is studied using a knowledge-based as well as a neural network based approach. The performance of the proposed features and approaches is studied on phones from databases of two languages, namely English and Hindi.

Efficacy of the discreteness of voicing category (DOVC) measure for characterizing voicing errors in children with cochlear implants: a report.

Abstract: Purpose This investigation explored the utility of an acoustic measure, called the discreteness of voicing category (DOVC), in identifying voicing errors in stop consonants produced by children wit...

Laryngeal adjustments in the production of consonant clusters and geminates in Moroccan Arabic

Abstract: The laryngeal adjustments in voiceless consonant clusters and geminates in Moroccan Arabic were examined by means of photoglottography. Findings indicate that speech rate and word boundary have an effect not only on the shape of laryngeal abduction-adduction gestures but also on laryngealoral coordination.

On the Nature of Schwa-like Vocalic Elements within some Berber Clusters

Abstract: In Tashlhiyt Berber, phonetic implementation creates what looks like a schwa vowel but in fact is not, at least of the phonological sort. We present various arguments against the interpretation of these elements as epenthetic segments introduced by the phonological component. In addition to the fact that they do not interact with phonology, our acoustic and articulatory data suggest that these schwa-like elements do not have their own time slot and are not associated with a specific articulatory target. The presence of these vocoids is claimed to be related to the amount of overlap and the voicing specifications of adjacent consonants.

Comparing the 'MagnifyingLens' Effect of Stress to that of Contrastive Focus in Spanish

Abstract: Stress, which is the linguistic property that makes a syllable in a word more prominent than others, has been defined as ‘localized hyperarticulation’ (de Jong 2004). Hyperarticulation is a term coined by Lindblom (1990) that explains phonetic variation as a function of communicative and situational demands. Since speech production is adaptive, speakers make certain parts of the speech more intelligible to the listener regardless of the articulatory effort it may cost them because these hyperaticulated speech parts carry the most relevant information in the message. Defining stress as localized hyperarticulation implies that its function consists of conveying relevant linguistic information to the listener in a clearer manner. Thus, the effect of stress on the speech signal is that of a magnifying lens it amplifies contrasts. Focus, in particular contrastive focus, has a similar function to stress as it makes a word more prominent within the sentence. In stress-accent languages, the device used to convey focus on a word is the placement of a pitch accent on the stressed syllable of that word (Pierrehumbert & Beckman 1986). For example in the sentence ‘No, MARY’s coming’, spoken as an answer to ‘Is Peter coming?’, ‘Ma’, the stressed syllable of ‘Mary’, bears a pitch accent. By means of this intonational pattern the speaker attracts the listener’s attention to the focused word, and, consequently, the stressed syllable of the focused word is expected to be hyperarticulated (de Jong 2004). De Jong (2004) and de Jong and Zawaydeh (2002) provide evidence for stress and focus as a case of localized hyperarticulation. They show that stress enlarges the vowel durational differences that cue vowel contrasts, such as long and short vowels in Arabic or American /e/ and /ae/. However, the vowel durational differences that cue the voicing contrast of a following stop are enlarged only in English, because Arabic uses other phonetic parameters to cue voicing. In addition, when the stressed syllable is placed under focus, those contrasts that have been magnified under stress become further amplified. Based on these results, de Jong (2004) concludes that the magnifying lens effect of stress and focus consists of amplifying only those differences relevant to the phonology of a particular language. Hay, Momoto, Coren, Moran & Diehl (2006) tested the above hypothesis in English, French and German by comparing spectral and durational differences in stressed vowels that were placed in focus and non-focus contexts. Their results indicate that, in spite of their different phonological systems, the three languages resort to the same strategy when amplifying spectral differences under focus: all vowels expand their acoustic/auditory space. However, languages differ in the way duration is used. Only German speakers enlarge durational differences since in German there is a contrast between long and short vowels. They conclude that it is difficult to confine the amplifying effect of focus to the relevant phonological contrasts of a language. Since a contrast can be cued by several acoustic parameters that enhance each other auditorily, and a single parameter can contribute to the enhancement of different contrasts (Kingston & Diehl 1994), it is difficult to decide which parameters are the most relevant in a language. Instead, the authors propose that “the amount of contrast enhancement of a vowel property in [+focus] context is possibly related to the between-category variation of that property in [-focus] context.” (Hay et al. 2006: 3031). In summary, it is clear that stress and focus have a magnifying lens effect on some phonetic material; however, which phonetic material is being modified and why, and whether this material is identical in stress and focus across languages are questions to be investigated here. Specifically, this

Voicing control and nasalization

Abstract: The dependency between nasality and voicing has been accounted for by ‘redundancy rules’ or OT constraints (e.g., NAS/VOI or *NC8), amongst others. This paper presents an account of the physical factors responsible for the dependency between nasality and voicing. In particular, it provides an aerodynamic explanation for cross-linguistic patterns that cannot be accounted for by acoustic-auditory factors (e.g., the lesser tolerance of voiceless stops vis-a-vis voiced stops to nasalization; Ohala and Ohala, 1991), the *NC8 constraint (Pater, 1999), or post-nasal voicing (Hayes and Stivers, 1996). For example, languages with distinctive voiceless stops, [p t k], and prenasalized voiced stops, [b, d, g], (illustrated in 1) but no simple voiced stops (suggesting that such prenasalized stops form the voiced stop series); phonetic preand postnasalization of voiced (but not voiceless) stops in the absence of contextual nasals (illustrated in 2); the emergence of non-etymological nasals adjacent to voiced but not voiceless stops (see 3) cannot be accounted for by the lesser tolerance of voiceless stops to nasalization or by postnasal voicing simply because these cases do not contain a nasal etymologically or occur in a nasal context. Similarly, the maintenance of the voicing stop contrast exclusively in a nasal context (see 4) cannot be explained by the principles above. (1) Voiceless vs prenasalized stops, e.g., Waris [b] banda ‘snake’ [p] panda ‘pitpit type’ [ŋg] go ‘go’ [k] kao ‘tree sp.’ (2) Prenasalization, e.g., Bola [b] ∼ [mb] bahele 'crocodile’ [d] ∼ [nd ] dagi 'dig' [g] ∼ [ŋg] ge '3 pers. fem' Post-nasalization, e.g., Lancashire [ˈspɪt ə ˈgɒbm] spit a gob [ˌu:z ə ˈwɛdn] she’s wed [ˈkɔ: ə ðɪ ˈlɛgŋ] calf of thy leg (3) Child phonology: English [bent] for bed [bed]; Greek [ŋgol] for [gol] ’goal’ Second language acquisition: Spanish [baɲo] for bano [baɲo] ‘bathroom’ (4) Stop voicing contrast preserved only postnasally, e.g., Majorcan Catalan /b/ dobl [pl] ‘I double’ sembl [bl] ‘I think’ /p/ acopl [pl] ‘I fit together’ umpl [pl] ‘I fill’ Basaa pen ‘color’ li-pem ‘honor’ m-pen ‘prong of a fork’ lep ‘water’ – – m-ben ‘handle’ –