Nasal Phonemes Research Articles

Personalized Audio-Driven 3D Facial Animation via Style-Content Disentanglement.

We present a learning-based approach for generating 3D facial animations with the motion style of a specific subject from arbitrary audio inputs. The subject style is learned from a video clip (1-2 minutes) either downloaded from the Internet or captured through an ordinary camera. Traditional methods often require many hours of the subject's video to learn a robust audio-driven model and are thus unsuitable for this task. Recent research efforts aim to train a model from video collections of a few subjects but ignore the discrimination between the subject style and underlying speech content within facial motions, leading to inaccurate style or articulation. To solve the problem, we propose a novel framework that disentangles subject-specific style and speech content from facial motions. The disentanglement is enabled by two novel training mechanisms. One is two-pass style swapping between two random subjects, and the other is joint training of the decomposition network and audio-to-motion network with a shared decoder. After training, the disentangled style is combined with arbitrary audio inputs to generate stylized audio-driven 3D facial animations. Compared with start-of-the-art methods, our approach achieves better results qualitatively and quantitatively, especially in difficult cases like bilabial plosive and bilabial nasal phonemes.

Problemy z interpretacją fonologiczną samogłosek nosowych w języku polskim

PROBLEMS WITH PHONOLOGICAL INTERPRETATION OF NASAL VOWELS IN POLISH The article discusses problems related to the sound realisation of the nasal sounds “ą” and “ę” as well as the differences in the phonological interpretation of those realisations. The issue has been raised many times by linguistic researchers, which proves its complexity. The article is an overview as its main goal has been to compare suggestions in terms of the nasal phonemes range, including the author’s suggestion for the phonological interpretation of the second part of the biphonemic sound production of the sounds “ą” and “ę” using phonological opposition and sound distribution.

The Velar Nasal in thing and think : Evidence from Thomas Spence's (1775) Pronouncing Dictionary for the Eighteenth-Century English Phonology Database

ABSTRACT: The history of the velar nasal phoneme involves two stories: the emergence of /ŋ/ as an independent phoneme, as in thing /θiŋ/, and variation between velar and alveolar nasal pronunciations in the context of unstressed <-ing>, as in loving /'lΛviŋ/–/'lΛvin/. The latter has been the object of study in English (historical) phonology and dialectology (e.g., , , , ). The former, however, received little attention until the final decades of the eighteenth century, when Thomas Spence first coded the velar nasal with a separate notation system in The Grand Repository of the English Language (1775). Spence's is one of the eleven pronouncing dictionaries from which the Eighteenth-Century English Phonology Database ( ECEP ) draws its data. The current paper reports on the methodological steps adopted in order to compile a new consonant lexical set for the study of the velar nasal phoneme /ŋ/, labeled thing and further subdivided into two subsets to account for the combination of the nasal consonant spelled <n> and a voiced /g/ or voiceless velar plosive /k/ consonant ( thing and think ). The selection of example words is based on the inventory of dictionary entries in The Grand Repository , then contrasts Spence's entries with other pronouncing dictionaries in ECEP and narrows down the list according to diverse linguistic parameters such as word-position, stress, part of speech, and number of syllables. The compilation of the lexical set thing will thus facilitate research on the incidence and distribution of the velar nasal phoneme in the history of English. This in turn will add to the body of research underlining the value of evidence drawn from pronouncing dictionaries and will contribute to fostering the "phonological turn" that has been growing in historical linguistics over recent decades.

Nasal phonemes and their realisation in Polish. Selected teaching problems

This article discusses the problems related to the sound realisation of the so-called nasal sounds, that is letters ą and ę, as well as the differences in the phonological interpretation of those realisations. The issue has been repeatedly raised by linguists (e.g. Zarembina 1963; Retz 1979ab; Dunaj 1991; Sawicka 1995; Nagórko 1998; Szpyra-Kozłowska 2002, and others), which proves its complexity and difficulty. The letters ą and ę can be pronounced in various ways depending on their neighbourhood (that is other letters that follow them), which causes problems in Polish language didactics and in the process of teaching of the correct Polish language pronunciation (especially when it comes to teaching Polish pronunciation to foreigners). The article mentions the methods and principles of the realisation of nasal sounds as well as inconsistencies in the phonological characteristics of the phenomenon. The article is an overview, as its aim is to compare different suggestions for the phonological interpretation. Keywords: phonetics – phonological system – nasal vowels – nasal phonemes.

Phonetics and phonology of nasality in Ecuadorian Siona

This study describes the nasal system in Ecuadorian Siona, an endangered Western Tukanoan language spoken in the Ecuadorian province of Sucumbíos, using the Earbuds Method to analyze nasal events acoustically. This method provides a visual representation of the timing and duration of velum gestures through intensity (dB) and amplitude (Pa) fluctuations in the nasal and oral cavities. The studied events include nasal spreading (nasal harmony), triggers, targets, blockers, and transparent segments. Meanwhile, differences between nasal phonemes and nasal allophones are also identified along with the effects of morpheme boundaries during nasal spreading events. Results reveal that, unlike many other Tukanoan languages, /m/ and /n/ function as individual phonemes independent of their oral counterparts (/p̰/ & /t̰/). In addition, nasal harmony was identified as predominantly rightward spreading apart from syllable-delimited leftward spreading to vocoid segments. Moreover, suffixes responsible for blocking nasal spreading appear to be reminiscent of oral suffixes in Eastern Tukanoan languages. Finally, more blockers were identified in Ecuadorian Siona than in most Eastern Tukanoan languages.

Voicing in Qaqet: Prenasalization and language contact

Qaqet is a non-Austronesian Baining language of Papua New Guinea, with a very small phoneme inventory of 16 consonants and four vowels, including the voiced stops /b d ɡ/. These stops are often phonetically realized as prenasalized [mb nd ŋɡ], and this feature is assumed to be a result of language contact with surrounding Oceanic languages. Our data consist of isolated word recordings from six female speakers of the language. Using a range of acoustic measures, we compare these phonetically prenasalized stops with non-prenasalized tokens of the same /b d ɡ/ phonemes, and with the nasal phonemes proper /m n ŋ/, as well as with the unaspirated voiceless stops /p t k/.In general we find that the nasal murmur of the phonetically prenasalized stop does not fully resemble the nasal murmur of the nasal consonant proper – instead the phonetically prenasalized stop patterns between the nasal consonant and the phonetically non-prenasalized voiced stop. However, there is a clear place effect, whereby the phonetically prenasalized velar stop patterns more closely with the nasal consonants, and the phonetically prenasalized bilabial stop patterns more closely with the phonetically non-prenasalized voiced stops – with phonetically prenasalized alveolar in between the bilabial and the velar. This is particularly reflected in the distribution of energy below and above about 350 Hz. However, measures of voicing strength suggest that voicing for the velar is weaker across all manners of articulation, in line with the general difficulty of maintaining voicing at this place of articulation. We conclude that prenasalization of the voiced stops largely serves to maintain voicing for the velar place of articulation; and that if the feature of prenasalization was borrowed from neighbouring languages, it was to maintain voicing for long closure durations in a true voicing language, particularly at places of articulation where maintaining voicing is difficult.

Selective Application of Pronouncing Strategies in the Production of Sequence of English Liquids by Elementary EFL Learners

This study aims to investigate how Korean elementary EFL learners pronounce English liquids in various contacts across words and see how they utilize two principles of selective application in the acquisition of English liquids concerning gender difference. For this, the study examines the liquids in the onset of the following word which contacts the coda of the preceding word with liquid or nasal phoneme. In the production test, 102 Korean elementary school students read 8 sentences and their recorded data are statistically analyzed by one-way ANOVA and three-way mixed ANOVA. The results are as follows. First, there is no significant difference between genders in the overall results of the English liquid production test. Second, despite no significant difference in accurate production, a significant main effect is reported in a repeated mixed ANOVA concerning error types of each contact context. Third, on the basis of pronouncing strategies, the ‘preservation principle’ and L1 phonological rule are differently adapted according to their contact contexts.

Acoustic examination of the duration of nasal phonemes in morae in people with profound hearing impairment

Acoustic examination of the duration of nasal phonemes in morae in people with profound hearing impairment

Is the Cepstral Analysis Sensitive Enough to Detect Untrained/Trained Resonant Voice in Healthy Subjects? A Preliminary Study

Objectives: In training resonant voice, clinicians rely on listening and feeling sensations that are subjective in nature. In the present study, the aim was to find out if cepstral analysis was sensitive enough to distinguish a resonant voice from an unresonant voice in healthy subjects. It was suggested that the immediate effect of humming exercise could be shown by accompanying higher Cepstral Peak Promince (CPP) and lower CPP standard deviation (CPP SD). Materials and Methods: Study included 20 vocally healty subjects aged between 20-30 years. All the participants were trained Humming exercise by teaching airflow technique. The recording protocol was repeated before and after resonant voice training. The recordings were gathered for the three samples of a) Humming b) A sentence weighted with nasal phonemes and c) A sentence weighted with all- voiced phonemes. Cepstral analysis recordings were gathered by the ‘Kay Elemetrics Group CSL Model 4300B’ equipment via the Analysis of Dysphonia in Speech and Voice software. Results: In both sentences, CPP values were found numerically higher and CPP SD values were found to be numerically lower after training. The difference was found statistically significant for the post-training value of voiced- weighted sentence (p=0.004). Conclusion: The higher numerical CPP values and lower CPP SD values for both sentences after resonant voice training indicated that cepstral measures were sensitive to detect trained/untrained resonant voice. However, these preliminary findings should be investigated in a larger sample size in order to achieve more reliable data.

Математичні метод і модель диференціації фоностатистичних структур авторського стилю

Розроблено метод комплексного аналізу диференціації фоностатистичних структур авторського стилю англійської мови, який ґрунтується на поєднанні трьох статистичних критеріїв: критерію Стьюдента, критерію Колмогорова-Смірнова і критерію хі-квадрат. Поєднання цих критеріїв дає змогу підвищити достовірність диференціації авторських стилів. Для розв'язання задачі диференціації авторських стилів побудовано статистичну модель, яка підвищує достовірність результатів авторської атрибуції тексту. Розроблена програмна система реалізує метод і модель з використанням мови програмування Java, що забезпечує платформонезалежність. Для тестування програми вибрано статті С. Логан і Д. Вебстер з газети "Вільна газета" ("Freedom Paper", papers by S. Logan and D. Webster). Істотні відмінності встановлено за групами носових, дорсальних і велярних фонем за критерієм Стьюдента, за всіма вісьмома групами фонем за критерієм Колмогорова-Смірнова, істотні відмінності встановлено за групами сонорних, щілинних, дорсальних, зімкнених, губних і носових фонем за критерієм хі-квадрат. Поєднання використаних критеріїв дало змогу встановити групу фонем з найбільшою авторорозрізняльною здатністю. Це група дорсальних фонем. За цією групою можна диференціювати тексти різних авторів, що дає змогу змінімізувати кількість груп фонем, за якими здійснюється авторська атрибуція тексту.

Метод комплексного аналізу диференціації фоностатистичних структур стилів англійської мови

Розроблено метод комплексного аналізу диференціації фоностатистичних структур стилів англійської мови. Метод ґрунтується на поєднанні двох статистичних критеріїв перевірки гіпотези на однорідність вибірки: критерію Стьюдента і критерію Колмогорова-Смірнова. Поєднання даних критеріїв забезпечує підвищення ефективності диференціації стилів. На основі розробленого методу, побудовано статистичну модель визначення ступеня дії чинника авторської манери викладу. Модель дає змогу підвищити ефективність стильової та авторської атрибуцій тексту. Розроблені метод і модель програмно реалізовано мовою програмування Java. POST запити двох типiв: /process i /process/transcription. розроблено у програмi. За відсутності транскрипційного варіанта досліджуваного тексту, використовується перший запит, другий – за його наявності. Скоротити час роботи програми дає змогу другий запит. Вбудовану базу даних H2 написано мовою програмування Java. База даних H2 є відкритою, кросплатформною. Вона підтримує мову SQL, має добру інтеграцію із використовуваним фреймворком Spring Boot i не потребує додаткових інсталяцій. У структурі даних HashMap зберігається відповідь із сайту транскрипцiйного перекладу. Структура побудована на принципі ключ-значення i дає змогу уникати дублікатів. Якщо опрацьовується велика кількість даних, то зменшується кількість запитів у мережі Інтернет, що забезпечує незалежність і автономність програми. Малі затрати часу є характерними для роботи програми.

V-Speech

Smart glasses are often used in public environments or industrial scenarios that are relatively noisy. Background noise and sound from competing speakers deteriorate voice communication or performance of automatic speech recognition (ASR). Typically, signal processing techniques are used to reduce noise and enhance voice quality, but they have limitations in performance, hardware and/or computing resources. Voice capturing techniques using bone conducting on the head have been proposed in some experimental and commercial devices, with good robustness against environmental noise, but limited by signal distortions inherent to the capturing method. We present V-Speech, a novel sensing and signal processing solution that enables speech recognition and human-to-human communication in very noisy environments. It captures the voice signal with a vibration sensor located in the nasal pads of smart glasses and performs a transformation to the sensor signal in order to mimic that of a regular microphone in low noise conditions. The signal transformation is key, as it eliminates the "nasal distortion" that is introduced for nasal phonemes in the speech induced vibrations of the nasal bone. The output of V-Speech has low noise, sounds natural, and can be used in voice communication or as input to an off-the-shelf ASR service. We evaluated V-Speech in noise-free and noisy conditions with 30 volunteer speakers uttering 145 phrases and validated its performance on ASR engines and with assessments of voice quality using the Perceptual Evaluation of Speech Quality (PESQ) metric. The results show in extreme noise conditions a mean improvement of 50% for Word Error Rate (WER), and 1.0 on a scale of 5.0 for PESQ. In addition, real life recordings were made under various representative noise conditions, some with sound pressure levels of 93 dBA, which require hearing protection. Subjective listening tests were conducted according to a modified ITU P.835 approach to determine intelligibility, naturalness and overall quality. Under these extreme conditions, where V-Speech achieved 30 dB SNR, subjective results show the speech is intelligible, and the naturalness of the speech is rated as fair to good. This enables clear voice communication in challenging work environments, for example in places with industrial, factory, mining and construction noise. With our proposed smart switching technique between a regular microphone signal and V-Speech, the optimal quality can be maintained from low to high noise conditions.

Sequence of Nasal-Stops as a Distinct Segment in Balinese

The article entitled “Sequence of Nasal-Stops as A Distinct Segment in Balinese” attempts to investigate wheather Balinese has sequence of nasal stops both bilabial nasal stops (<m-b>, <m-p>), alveolar nasal-stops (<n-d>, <n-t>), and velar nasal-stops (<ng-g>, <ng-k> or it has a prenasal stops both bilabial prenasal stops (<mb>, <mp>), alveolar prenasal stops (<nd>, <nt>) or it has velar prenasal stops (<ngg>, <ngk>). Based upon the data analysis, it has sequence of nasal-stop segments. The judgement is based on (1) the existence of suspicious syllabic nasal phonemes, (2) its clear word structure such as (a) VC.CVC e.g. <aggag> /aggag/ [ag.gag] ‘open’, (b) CVC.CV e.g. <bakta> /bakta/ [bak.ta] ‘bring’, (c) CVC.CVC e.g. <dapdap> /dapdap/ [dap.dap] ‘careful’ and (d) VC.CV e.g. <akta> /akta/ [ak.ta] ’official document’. (3) On the bases of its clear word structure, Balinese shares sequence of nasal-stop segments both bilabial nasal-stop such as <m-b> e.g. <embok> ‘elder sister’ of which the right pronunciation is [ǝm.bɔk] rather than prenasal stop [ᵐb] as in [ᵐbɔk], alveolar nasal-stop such as <n-d> in <cande> ‘just for kidding’ of which the right pronunciation is [can.dǝ] rather than prenasal stop [ⁿd] in [ca.ⁿdǝ] and velar nasal-stop such as <ng-g> in <bengkek> ‘short body posture’ of which the right pronunciation is [bǝŋ.kǝk] rather than velar prenasal [ᵑg] in [bǝ.ᵑkǝk].

Speaker recognition based on Arabic phonemes

In this paper, we investigate the effect of Arabic phonemes on the performance of speaker recognition systems. The investigation reveals that some Arabic phonemes have a strong effect on the recognition rate of such systems. The performance of speaker recognition systems can be improved and their execution time can be reduced by utilizing this finding. Additionally, this finding can be used by segmenting the most effective phonemes for speaker recognition from the utterance, using only the segmented part of the speech for speaker recognition. It can also be used in designing the text to be used in high-performance speaker recognition systems. From our investigation, we find that the recognition rates of Arabic vowels were all above 80%, whereas the recognition rates for the consonants varied from very low (14%) to very high (94%), with the latter achieved by a pharyngeal consonant followed by the two nasal phonemes, which achieved recognition rates above 80%. Four more consonants had recognition rates between 70% and 80%. We show that by utilizing these findings and by designing the text carefully, we can build a high-performance speaker recognition system.

Quantitative Endoscopic Phototransducer Investigation of Normal Velopharyngeal Physiology.

The purpose of this research was to learn the extent to which healthy individuals vary in their ability to achieve velopharyngeal closure for speech. Twenty healthy adult volunteers (10 women, 10 men) were tested using an endoscopic phototransducer system that tracks variations in velopharyngeal closure during speech production. Each speaker produced multiple repetitions of three utterances that differed in phonetic content. The data were amplitude normalized and averaged for each speaker. Average phototransducer measurements were similar across subjects for utterances containing only oral phonemes. Average percentage of velopharyngeal closure varied considerably among subjects when producing utterances containing both oral and nasal phonemes (54%-95%). Average percentage of velopharyngeal closure levels were significantly lower (p < .05) for utterances that included nasal consonants. Phototransducer measurements of velopharyngeal closure for speech are sensitive to nasal phoneme content. The findings suggest that motor programming that accomplishes rapid oral-nasal velopharyngeal valving for speech may differ among healthy subjects. However, such variations in motor programming may not perceptually affect typical speakers. If present in individuals with abnormal velopharyngeal mechanisms, these variations may help explain variations among speakers in speech outcomes after physical and behavioral management.

The status of voiceless nasals in Miyako Ryukyuan

The current study investigates voiceless nasals in Miyako, a language spoken on remote Japanese islands near Taiwan, focusing on the dialect Ikema. Voiceless nasals are rare cross-linguistically. Hayashi (2013) suggested that there is phonemic contrast between voiced /n, m/ and voiceless /n, m/. Pilot data from one speaker both confirmed and refuted Hayashi’s claims, analysis indicating that these voiceless nasals occur both as voiceless and breathy voiced. Thus, the glottal state of this phoneme may depend on the phonetic environment. The current study further explores these findings based on data collected from four male and two female speakers. Participants were asked to produce target and minimal pair words in isolation and within novel sentences. Voiceless and voiced nasal segments were compared and analyzed for duration, cepstral peak prominence, and amplitude measures of H1 and H2 to determine whether breathy voicing state might differentiate target phonemes from /n/ and /m/. The results of this analysis are used to investigate the nature of this phonemic contrast and how Ikema is situated cross-linguistically in its use of voicing to distinguish nasal phonemes.

Nen

Nen (ISO 639–3 code: nqn) is the easternmost language of the Yam (Morehead-Maro) family of Southern New Guinea. This family is one of over forty maximal genetic groupings in New Guinea, and is currently not relatable to any other language family in New Guinea or elsewhere. As with other aspects of their grammar, the phonology of the Morehead-Maro languages differs significantly from those of other Papuan languages. The phonology of Nen is broadly similar to that of other members of the family, except that (like most other members of the Nambu branch, of which it is a member) it lacks the velar nasal phoneme found in more westerly languages, and it has a smaller fricative inventory: other languages of the family include /ɸ ɣ θ/, all absent from Nen.

The identification of American English vowels by native speakers of Japanese before three nasal consonants

Native speakers of Japanese identified American English vowels that are uttered before three nasal consonants /m, n, ŋ/ and three oral stop consonants /b, d, ɡ/. Of the seven vowels /i, ɪ, eɪ, ɛ, æ, ɑ, ʌ/, /æ/ was generally less accurately identified before nasal consonants than before oral stop consonants, and this tendency was stronger when /ŋ/ follows. This tendency is probably attributed to the extended raising of /æ/ before /ŋ/ and the Japanese listeners’ limited sensitivity to differentiate three nasal phonemes in coda position. /ɪ/, on the other hand, was identified more correctly before /ŋ/ than before the other two nasal consonants, also probably because the vowel is raised before /ŋ/. This vowel was more often misidentified as /ɛ/ before /m/ and /n/. /ɑ/ and /ʌ/ were less accurately identified before stop consonants, but after nasal consonants, /ʌ/ was more often misidentified as /ɑ/. /ɑ/ and /ʌ/ may sound alike to Japanese listeners in every context, but before nasal contexts, both of these vowels may sound closer to the Japanese vowel /o/. The results generally revealed that identification accuracy cannot be solely accounted for in terms of the place of articulation of the following consonant.

Impact of Obturation of Palatal Fistulae on Speech Quality and Aerodynamic Parameters in Patients with Cleft Palates

Objectives: The aim of this prospective study was to measure nasal and oral airflow during speech, before and after obturation. Patients and Methods: Included were children aged 3-18 years with nonsyndromic clefts and palatal fistulae. The corpus used was: syllable /pi/; a sentence containing stop consonants and a nasal phoneme; and the description of a picture of a scene. Analysis criteria were: percentage of nasality; value of average flow for the explosion; perceived nasality and intelligibility; and tolerance of the proposed device. Results: Only 5 children were included due to the observation of an increase in the percentage of nasality after obturation. The value of average flow for the explosion increased in all patients. A decrease in perceived nasality was noted in all but 1 patient. An improvement in intelligibility was observed in 3 out of the 5 children. The tolerance of the device was good. Conclusion: While the small number of patients studied does not permit firm conclusions concerning the efficiency of the obturation, the method described, as well as the introduction of ‘speed of explosion' of stop consonants, offer new perspectives to prospectively study obturator effects on speech.

Speech Methodologies in Cleft Palate Assessment: A Pilot Project to Compare Two Phonetically Different Sentence Sets

Objective: To investigate the impact of the phonetic content of two sentence sets on speech outcomes, specifically the effects of nasal phonemes. Method: Audio-video recordings of a consecutive series of 15 participants (age range 4–22 years), with cleft palate (syndromic or non-syndromic), with and without velopharyngeal dysfunction were taken. Participants repeated Sentence Set 1 (with nasals across sentences) and Sentence Set 2 (without nasals except the three nasal target sentences) during a routine speech recording. Two experienced Specialist Speech and Language Therapists, blinded to the study’s purpose, analyzed participants’ speech using the Cleft Audit Protocol for Speech-Augmented (CAPS-A). On day 1, recordings included Sentence Set 1. On day 2, 23 days later, recordings included Sentence Set 2. Main results: The difference between Sentence Set 1 and Sentence Set 2 ‘total scores’ (sum of scores on all CAPS-A parameters) was significant. The Pearson Product Moment showed high correlation. A Wilcoxon test revealed a significant difference between Sets 1 and 2 on the hypernasality parameter, and this alone accounted for the significant difference in total scores. Conclusion: The inclusion or exclusion of nasal consonants in the sentence set significantly affected perceptual ratings of hypernasality but none of the other CAPS-A parameters, highlighting the need for further investigation into perceptual nasality ratings.