Articulatory Features Research Articles

Hierarchical Temporal Structuring of Speech: A Multiscale, Multimodal Framework to Inform the Assessment and Management of Neuromotor Speech Disorder.

Hierarchical temporal structuring of speech is the key to multiscale linguistic information transfer toward effective communication. This study investigated and linked the hierarchical temporal cues of the kinematic and acoustic modalities of natural, unscripted speech in neurologically healthy and impaired speakers. Thirteen individuals with amyotrophic lateral sclerosis (ALS) and 10 age-matched healthy controls performed a story-telling task. The hierarchical temporal structure of the speech stimulus was measured by (a) 26 articulatory-kinematic features characterizing the depth, phase synchronization, and coherence of temporal modulation of the tongue tip, tongue body, lower lip, and jaw, at three hierarchically nested timescales corresponding to prosodic stress, syllables, and onset-rime/phonemes, and (b) 25 acoustic features characterizing the parallel aspects of temporal modulation of five critical-spectral-band envelopes. All features were compared between groups. For each aspect of temporal modulation, the contributions of all articulatory features to the parallel acoustic features were evaluated by group. Generally consistent disease impacts were identified on the articulatory and acoustic features, manifested by reduced modulation depths of most articulators and critical-spectral-band envelopes, primarily at the timescales of syllables and onset-rime/phonemes. For healthy speakers, the strongest articulatory-acoustic relationships were found for (a) jaw and lip, in modulating stress timing, and (b) tongue tip, in modulating the timing relation between onset-rime/phonemes and syllables. For speakers with ALS, the tongue body, tongue tip, and jaw all showed the greatest contributions to modulating syllable timing. The observed disease impacts likely reflect reduced entrainment of speech motor activities to finer-grained linguistic events, presumably due to the dynamic constraints of the neuromuscular system. To accommodate these restrictions, speakers with ALS appear to use their residual articulatory motor capacities to accentuate and convey the perceptually most salient temporal cues underpinned by the syllable-centric parsing mechanism. This adaptive strategy has potential implications in managing neuromotor speech disorders.

Acoustic-articulatory emotion recognition using multiple features and parameter-optimized cascaded deep learning network

Multimodal emotion recognition is an important research direction within artificial intelligence. In this study, we propose a model for acoustic-articulatory emotion recognition. For the acoustic data, this model extracts Interspeech 2009 (IS 09) features, and for articulatory data, it extracts our proposed phase space reconstruction-geometric (PSR-G) features. It then feeds the spliced features into an improved sparrow search algorithm (ISSA)-cascaded deep learning (CDL) network to obtain the final recognition results. We propose the PSR-G features to reflect phase and geometric information, where the reconstructed phase-space signal of the articulatory data is plotted in three-dimensional space and geometric features based on distances and angles are extracted. We also propose the ISSA-CDL network for emotion recognition, in which the CDL network effectively merges acoustic and articulatory features, and fully leverages the advantages of the one-dimensional convolutional neural network, a multi-head self-attention mechanism, and double-layer bidirectional long short-term memory. Finally, we propose the ISSA, in which we use a tent map and the firefly algorithm to optimize the parameters of the CDL network to decrease the instability and randomness induced by subjective experience. We conducted experiments using the self-recorded STEM−E2VA database and obtained the following results: (1) PSR-G features lead to higher recognition accuracy for the articulatory data than other existing features. (2) The CDL network effectively merges bimodal features, and the ISSA effectively optimizes the parameters of the CDL network. (3) The final accuracy in acoustic-articulatory emotion recognition is 95.87 ± 0.29%, which is higher than that for acoustic features (81.16 ± 0.47%) or articulatory features (93.27 ± 0.47%) alone.

Self-organizing speech recognition that processes acoustic and articulatory features

Self-organizing speech recognition that processes acoustic and articulatory features

Cross-linguistic variability in audiovisual enhancement of coronal consonants

The role of auditory perceptibility in phonological contrast is uncontroversial. In clear speech, speakers increase acoustic distance between contrastive phones, thereby improving perception. Although visual cues are also known to improve perception, it is unclear whether speakers actively enhance their speech for visual perceptibility. 20 native English and 20 native Mandarin speakers participated in a two-part production experiment. Words containing /l d n s r/ (and English /θ/) were elicited, with vowels balanced for height, backness, and rounding. Speakers first recited each word while alone in a soundproof booth. Words were then spoken for a native listener (who attempted to guess the word) over Zoom with heavily degraded audio. Audio and high-speed video were recorded. English speakers showed a significant increase in visually exaggerated articulations for /l, θ/ (but not /n, d, s, r/), including interdental and linguolabial productions with external lingual protrusion. Mandarin speakers increased their use of interdental articulation for /n, l, t, s/ (but not /r/), although lingual protrusion was limited. These results demonstrate that speakers may enhance their speech for visual over auditory perceptibility but suggest that enhancement is mediated by language-specific factors including phonological contrast and the articulatory characteristics of sounds in normal speech.

Articulatory working space in first and second languages

Articulatory working space (acoustic and kinematic) is often studied to understand the overall size (limits) of a speaker’s articulatory behaviors. For example, prior research has shown that the magnitude of articulators’ movement (e.g., maximum tongue advancement, lip aperture) changes as a function of speech effort (loud, clear, and slow speech). To better understand second language acquisition in adults (i.e., articulatory working space determined by the language or anatomical differences), we compare both acoustic and kinematic working space of adult learners of English between their first language (L1) and second language (L2), which is also compared with that of native speakers of English. Specifically, the articulatory convex hull is measured during passage reading for both acoustic (F1 and F2 trajectories) and kinematic (tongue trajectories on x- and y-dimensions) data. Participants include 11 adult learners of English (four men and seven women) with a Korean language background and 10 adult L1 speakers of English (six men and four women). In the presentation, the findings will be discussed to address whether articulatory space is (1) different between native and nonnative languages (determined by linguistic needs) or (2) rather a constant articulatory characteristic of speakers between the languages, regardless of the speaker’s English proficiency.

Amodal phonological abstraction in infants at 10 months

Infants can abstract amodal information about place of articulation at 4-5 months, before perceptual attunement to native consonants occurs (Altuntas et al., in preparation). We tested infants in the same task at 10 months, after native consonant attunement has commenced, which may affect phonological abstraction relative to 4–5 months. 29 infants acquiring Australian English were trained to associate different cartoon animals with audio-only words from two artificial mini-languages (Language A: Kangaroo; Language B: Kookaburra), which differed in use of only labial or only coronal sets of consonants. They were tested with video-only words (silent talking face) paired with Congruent (matching) or Incongruent (mismatching) animals. We hypothesized that infants would look longer at the Congruent than Incongruent pairings during the test trials, indicating generalization from the auditory (training) to the visual modality (test), i.e., amodal abstraction. Surprisingly, looking times did not differ between Congruent and Incongruent trials. Attunement to native consonants at 10 months may interfere with phonological abstraction of the place of articulation feature across different consonants that share the same place. Examining other phonological abstractions or comparing infants acquiring languages with different consonant inventories could shed light on how perceptual attunement affects amodal phonological abstraction.

An Analytical Study of Assimilation in English Reading Texts

Abstract This study is concerned with analysing of assimilation as an English co-articulated phenomenon in a selected reading course assigned for Iraqi students in order to enhance the Iraqi ESL learners understanding of variety of English reading skills as well as phonology which is one of the bases of effective and successful reading. The aim of this paper is to help students overcome their problems through reading, listening and fluently pronouncing assimilated words in connected speech of English utterances for successful interaction and minimizing the Iraqi ESL students’ foreign accent. The collected data are the soundtracks of 14 reading passages taken from “Intermediate-Select Readings” /Second Edition by Linda Lee & Erik Gundersen as a second year reading textbook at the College of Languages / University of Kufa. The study analysed the data under scrutiny and found that the reciprocal assimilation is not as it is defined as adjacent sounds having mutual influence on each other to produce a third new fused sound, some features are copied just by one of them. It was also revealed that germination mostly occurs in complete assimilation when consonants in sequence are homorganic (the consonants have two features in common, manner of articulation and phonation). Since manner of articulation feature is existed in both sounds, the third place of articulation feature is copied automatically from the adjacent sound. Keywords: Assimilation, Germination, Voicing, Place, Manner, Regressive, Progressive

Akustyka, mimetyka, estetyka, artykulacja: Grecy o głoskach języka

The paper contains a thorough examination of the criteria used by the ancient Greeks for defining, describing and classifying the phones of the language. References are made to works by Plato and Aristotle, as well as to the text of the Techne grammatike attributed to Dionysius Thrax. In the dialogues of Plato, and especially in the Cratylus, one can find the first categorical characteristics of different groups of phones, formulated in acoustic terms, as well as individual descriptions of the mimetic qualities of particular phones, based on either sound effects or articulatory features. In Poetics, Aristotle defined the phone as such, connecting it with the notion of voice characterized by indivisibility, ability to create a complex voice (i.e. a word) and exclusively human origin. In setting out three main classes of phones, he based their definitional criteria on their articulatory features (the contact of speech organs or lack of it) on the one hand, and on their phonetic and acoustic properties (audible voice), i.e. on sound distinctness manifesting itself in acoustic autonomy on the other. In the Techne grammatike, in turn, particular classes of phones were identified and defined by formulating the conditions in which they produce voice characterized by a definite – high or law – aesthetic value. Despite the absence of an articulatory criterion, Greek grammarians succeeded in distinguishing all basic classes of phones, although they constantly looked at phones through the prism of letters, which led to them confusing the phonetic level with the graphic one, and thus made it impossible to create a conceptual and terminological apparatus which would be the means appropriate for identifying and describing correctly the whole stock of Greek phones.

OUR EXPERIENCE IN THE USE OF BONE-CARTILAGE AUTOPLASTY IN THE TREATMENT OF PATIENTS WITH OSTEOCHONDRAL PATELLAR DEFECTS

Objective: to evaluate the effectiveness of bone-cartilage autoplasty in the treatment of patients with osteochondral patellar defects. Materials and methods. The study analyzed the results of surgical treatment of 7 patients for the period from 2012 to 2023 inclusive. All patients underwent inpatient treatment at the Department of Traumatology and Orthopedics No. 2 Clinics of Samara State Medical University (SamSMU). All patients had a defect of patellar cartilage on the background of osteoarthritis of the knee joint with a predominant lesion of the patellofemoral joint (patellofemoral arthrosis). Bone-cartilaginous autoplasty of the articular surface of the patella was used as an operation. Clinical evaluation of treatment results was carried out on the basis of the methods recommended by OARSI (Osteoarthritis Research Society International) before surgery and at 3, 6 and 12 months after surgery: VAS (Visual Analog Scale), WOMAC Index (Western Ontario and McMaster University Osteoarthritis Index) and M. Lequesne algofunctional index were determined. The condition of bone-cartilage tissue in the defect area before surgery and after bone-cartilage autoplasty was assessed using radiography and magnetic resonance imaging. Results and discussion. We noted that according to the studied indices and scales, the greatest positive dynamics was observed in the early postoperative period (3 months), the peak reached by 6 months, and full functional recovery occurred by 12 months (return to sports, high-intensity work). The most complete restructuring as a result of the examination of 7 patients according to radiography and MRI, we noted at the time of 6 months, when we see a complete restructuring of the patellar cartilage tissue, as well as the greatest signs of autograft integration in the subchondral zone. Conclusions. The use of bone-cartilage autoplasty as a method of surgical treatment in patients with osteochondral injuries of the patella in clinical practice is effective, expedient and its use in clinical practice is recommended from the standpoint of restoring the function of the knee joint, which leads to an improvement in the quality of life of patients. Nevertheless, its application requires taking into account all anatomical and functional features of the patellofemoral articulation.

Near-Optimal Active Learning for Multilingual Grapheme-to-Phoneme Conversion

The construction of pronunciation dictionaries relies on high-quality and extensive training data in data-driven way. However, the manual annotation of corpus for this purpose is both costly and time consuming, especially for low-resource languages that lack sufficient data and resources. A multilingual pronunciation dictionary includes some common phonemes or phonetic units, which means that these phonemes or units have similarities in the pronunciation of different languages and can be used in the construction process of pronunciation dictionaries for low-resource languages. By using a multilingual pronunciation dictionary, knowledge can be shared among different languages, thus improving the quality and accuracy of pronunciation dictionaries for low-resource languages. In this paper, we propose using shared articulatory features among multiple languages to construct a universal phoneme set, which is then used to label words for multiple languages. To achieve this, we first developed a grapheme−phoneme (G2P) model based on an encoder−decoder deep neural network. We then adopted a near-optimal active learning method in the process of building the pronunciation dictionary to select informative samples from a large, unlabeled corpus and had them labeled by experts. Our experiments demonstrate that this method selected about 1/5 of the unlabeled data and achieved an even higher conversion accuracy than the results of the large data training method. By selectively labeling samples with a high uncertainty in the model, while avoiding labeling samples that were accurately predicted by the current model, our method greatly enhances the efficiency of pronunciation dictionary construction.

Pronunciation challenges faced by Brazilian EFL learners

The present study aims to identify the most prominent pronunciation challenges faced by Brazilian learners in a language course affiliated with Universidade Federal do Pará, and discuss which errors seemed to pose the greatest issues for communication. This study proposes to describe the articulation features of English sounds in association with pronunciation, as well as present the importance of teaching the speaking skill with an attention to pronunciation. In order to reach its main objective, it was necessary to gather sufficient data by observing these students during their classes. Furthermore, the study goes on to confirm that the errors presented by the authors Mascherpe (1970), Lessa (1985), Lieff (1993), Nunes (1993) and Baptista (2001), also occur in the group of learners who were the object of this investigation, and the results suggest that the most prominent issues were caused by errors related to stress placement, vowel insertion and consonants. However, the errors which caused the most problems for communication happened on the suprasegmental features of language, which in this case is represented by the category of stress placement.

Automatic classification of the severity level of Parkinson’s disease: A comparison of speaking tasks, features, and classifiers

Automatic speech-based severity level classification of Parkinson’s disease (PD) enables objective assessment and earlier diagnosis. While many studies have been conducted on the binary classification task to distinguish speakers in PD from healthy controls (HCs), clearly fewer studies have addressed multi-class PD severity level classification problems. Furthermore, in studying the three main issues of speech-based classification systems – speaking tasks, features, and classifiers – previous investigations on the severity level classification have yielded inconclusive results due to the use of only a few, and sometimes just one, type of speaking task, feature, or classifier in each study. Hence, a systematic comparison is conducted in this study between different speaking tasks, features, and classifiers. Five speaking tasks (vowel task, sentence task, diadochokinetic (DDK) task, read text task, and monologue task), four features (phonation, articulation, prosody, and their fusion), and four classifier architectures (support vector machine (SVM), random forest (RF), multilayer perceptron (MLP), and AdaBoost) were compared. The classification task studied was a 3-class problem to classify PD severity level as healthy vs. mild vs. severe. Two MDS-UPDRS scales (MDS-UPDRS-III and MDS-UPDRS-S) were used for the ground truth severity level labels. The results showed that the use of the monologue task and the articulation and fusion of features improved classification accuracy significantly compared to the use of the other speaking tasks and features. The best classification systems resulted in a rate of accuracy of 58% (using the monologue task with the articulation features) for the MDS-UPDR-III scale and 56% (using the monologue task with fusion of features) for the MDS-UPDRS-S scale.

Intonation as a means of conveying an utterance in Komi prose

Introduction. The purpose of this paper is to study intonation features and means of conveying speech acts in Komi works of art. The voice as a source of information was not in the focus of the attention of Komi linguists, the issues of Komi intonation and the means of their reflection have not been studied. In this regard, the address to the general characteristics of how Komi prose writers reflect the intonation features of their characters’ speech is considered relevant in the Komi language and allows us to speak about a new method of studying the perception of phrasal intonation. Materials and Methods. The material for the study based on the examples from the works of modern fiction of Komi classics. The methodological basis of the study was a set of such linguistic methods as descriptive, perceptual analysis of phrasal intonation, linguistic analysis of the text, elements of structural and semantic analysis. Results and Discussion. The study of intonation allowed us to state a number of prosodic signs. The described speech of the characters of literary works was studied according to various parameters of acoustics: duration, intensity, pitch, timbre, tempo, rhythm; by means of non-speech sounds, through an indication of the features of articulation, diction; by similarity with the phenomena of the sounding world. It was found that the author’s remarks of works of art describe the speech volume of the characters in particular detail, the intensity of the sounding speech. Conclusions. The research material has shown that the language functions of intonation can be fully and consistently reflected not only in live sounding speech, but also in writing in the form of the speech of the literary hero of prose works. The functions and expressive possibilities of intonation are contained in the author’s designations of prosodic characteristics of speech. An important role is played by the individual style of the author, which is distinguished by sensitivity to the sound side of the speech of thier characters.

Neural and Behavioral Differences in Speech Perception for Children With Autism Spectrum Disorders Within an Audiovisual Context.

Reduced use of visible articulatory information on a speaker's face has been implicated as a possible contributor to language deficits in autism spectrum disorders (ASD). We employ an audiovisual (AV) phonemic restoration paradigm to measure behavioral performance (button press) and event-related potentials (ERPs) of visual speech perception in children with ASD and their neurotypical peers to assess potential neural substrates that contribute to group differences. Two sets of speech stimuli, /ba/-"/a/" ("/a/" was created from the /ba/ token by a reducing the initial consonant) and /ba/-/pa/, were presented within an auditory oddball paradigm to children aged 6-13 years with ASD (n = 17) and typical development (TD; n = 33) within two conditions. The AV condition contained a fully visible speaking face; the pixelated (PX) condition included a face, but the mouth and jaw were PX, removing all articulatory information. When articulatory features were present for the /ba/-"/a/" contrast, it was expected that the influence of the visual articulators would facilitate a phonemic restoration effect in which "/a/" would be perceived as /ba/. ERPs were recorded during the experiment while children were required to press a button for the deviant sound for both sets of speech contrasts within both conditions. Button press data revealed that TD children were more accurate in discriminating between /ba/-"/a/" and /ba/-/pa/ contrasts in the PX condition relative to the ASD group. ERPs in response to the /ba/-/pa/ contrast within both AV and PX conditions differed between children with ASD and TD children (earlier P300 responses for children with ASD). Children with ASD differ in the underlying neural mechanisms responsible for speech processing compared with TD peers within an AV context.

Speech and sexual dimorphism: the history of the issue in the light of modern theories of the origin of speech and experimental data

In this theoretical article, the authors turn to the analysis of studies explaining the origin and revealing the functions and meaning of human oral speech within the framework of an evolutionary approach. First of all, the authors are interested in the influence of sexual selection on the development of verbal skills in different age and gender groups. In this regard, the data of the history of the human life cycle are of particular interest, which indicate that the mastery of language as a tool for influencing and manipulating the opinions of others approaches an advanced level of proficiency only at the time of puberty, and speech, thus, begins to play an important role in intersex competition. Rapidly developing with the onset of puberty, the ability to performative vivid performances, to defend their position, as well as to conduct debates in public, apparently, can be guided by the forces of sexual selection, and therefore, they should manifest themselves more clearly in men than in women. It is assumed that verbal mastery can serve as a means of actively attracting attention for the purpose of self-promotion and improving the social status of the speaker. The methodological basis of our work was largely the analysis of ethnographic sources. In the article, the authors cite extensive ethnographic material confirming the connection of high social status with oratorical abilities. The authors also turn to the latest research that analyzes the differences in performative speech between men and women. Experimental work shows that in a conversation with an attractive woman, men tend to be more creative in choosing words and expressions. Some researches consider the connection of various aspects of speech with the attractiveness of the speaker, his masculinity and adaptability. Of great interest are studies studying the relationship between musical abilities and attractiveness, as well as works analyzing the articulatory features of oral speech and their connection with belonging to a certain social environment or social class. All the features of speech given in the review could most likely have been formed by the pressure of the forces of sexual selection. In this regard, the study of sexual dimorphism seems to be one of the most important directions in the research of performative speech.

CNN-Based Identification of Parkinson's Disease from Continuous Speech in Noisy Environments.

Parkinson's disease is a progressive neurodegenerative disorder caused by dopaminergic neuron degeneration. Parkinsonian speech impairment is one of the earliest presentations of the disease and, along with tremor, is suitable for pre-diagnosis. It is defined by hypokinetic dysarthria and accounts for respiratory, phonatory, articulatory, and prosodic manifestations. The topic of this article targets artificial-intelligence-based identification of Parkinson's disease from continuous speech recorded in a noisy environment. The novelty of this work is twofold. First, the proposed assessment workflow performed speech analysis on samples of continuous speech. Second, we analyzed and quantified Wiener filter applicability for speech denoising in the context of Parkinsonian speech identification. We argue that the Parkinsonian features of loudness, intonation, phonation, prosody, and articulation are contained in the speech, speech energy, and Mel spectrograms. Thus, the proposed workflow follows a feature-based speech assessment to determine the feature variation ranges, followed by speech classification using convolutional neural networks. We report the best classification accuracies of 96% on speech energy, 93% on speech, and 92% on Mel spectrograms. We conclude that the Wiener filter improves both feature-based analysis and convolutional-neural-network-based classification performances.

MSilent

Silent speech recognition (SSR) allows users to speak to the device without making a sound, avoiding being overheard or disturbing others. Compared to the video-based approach, wireless signal-based SSR can work when the user is wearing a mask and has fewer privacy concerns. However, previous wireless-based systems are still far from well-studied, e.g. they are only evaluated in corpus with highly limited size, making them only feasible for interaction with dozens of deterministic commands. In this paper, we present mSilent, a millimeter-wave (mmWave) based SSR system that can work in the general corpus containing thousands of daily conversation sentences. With the strong recognition capability, mSilent not only supports the more complex interaction with assistants, but also enables more general applications in daily life such as communication and input. To extract fine-grained articulatory features, we build a signal processing pipeline that uses a clustering-selection algorithm to separate articulatory gestures and generates a multi-scale detrended spectrogram (MSDS). To handle the complexity of the general corpus, we design an end-to-end deep neural network that consists of a multi-branch convolutional front-end and a Transformer-based sequence-to-sequence back-end. We collect a general corpus dataset of 1,000 daily conversation sentences that contains 21K samples of bi-modality data (mmWave and video). Our evaluation shows that mSilent achieves a 9.5% average word error rate (WER) at a distance of 1.5m, which is comparable to the performance of the state-of-the-art video-based approach. We also explore deploying mSilent in two typical scenarios of text entry and in-car assistant, and the less than 6% average WER demonstrates the potential of mSilent in general daily applications.

Hadza liquid production and acoustics

Liquid systems in the world’s languages are relatively poorly understood. This study employs 3D/4D ultrasound methods to investigate articulatory characteristics of Hadza liquids produced by a single male speaker in a laboratory, analyzing simultaneously collected acoustic signals to determine relations between articulation and acoustics. Articulatory-acoustic relations are then compared to acoustic data obtained from other speakers collected in the field. Hadza is a language isolate spoken by approximately 1200 people (Brian Wood, pc.) in north-central Tanzania. Hadza is reported to have a single liquid phoneme which varies allophonically between a lateral approximant [l] occurring word-initially, and a tap [ɾ], occurring intervocalically (Tucker et al., 1977; Sands et al., 1996; Sands, 2013). The results of the current study confirm the production of both lateral approximant [l] and tap [ɾ] allophones; however, the results also show substantial variation between speakers. Furthermore, some speakers exhibit a categorical distinction between the allophones in given phonological environments, while others indicate a very similar articulation in both, and demonstrate no such distinction.

Lip posture of high and high-mid vowels in Hnaring Lutuv

Lutuv is an under-documented South Central Tibeto-Burman language which is reported to contain six high monophthongs /i, y, ɨ, ʉ, Ɯ, u/ and four diphthongal vowels with high onsets /ie̯, yə̯, uɔ̯, Ɯə̯/ (Bohnert etal., 2022). No articulatory work on Lutuv exists to date, but impressionistic observations suggest that while the contrast between /i, y/ and /u, Ɯ/ is one of rounding (as implied by their IPA transcriptions), the contrast between /ɨ, ʉ/ is instead one of lip spreading. In addition, /ʉ/ seems to involve a degree of frication from labiodental contact. To further investigate these observations, the current paper utilizes facial landmark recognition through OpenFace 2.0 (Baltrusaitis et al., 2018) to analyze the lip posture of these vowels, measuring horizontal, vertical, and lateral lip aperture over the duration of each vowel produced by one native speaker in a curated word list. These measurements are compared both within and across pairs to determine (1) the articulatory characteristics of Lutuv rounding contrasts, and (2) whether the central high vowels display these characteristics. These data complement existing acoustic analyses and provide a more detailed understanding of the articulation of a typologically rare contrast.

A preliminary study on laryngeal and supralaryngeal articulatory distinction of the three-way contrast of Korean velar stops*

This study investigated acoustic (VOT) and articulatory characteristics of Korean velar stops in monosyllabic CV structures to examine how the three-way distinction is realized in the laryngeal and supralaryngeal domains and how the distinction is manifested in male versus female speakers’ speech production. EMA data were collected from 22 speakers. In line with previous studies, male speakers preserved the three-way differentiation of velar stops (/k^*/</k/</k^h/) in terms of VOT while female speakers showed only a two-way distinction (/k^*/</k/=/k^h/). As for the kinematic characteristics, a clear three-way distinction was found only in male speakers’ peak velocity measure in the C-to-V opening movement (/k^h/</k/</k^*/). For the other kinematic measures (i.e., articulatory closure duration, deceleration duration of the opening movement and the entire opening movement duration), male speakers showed only a two-way distinction between fortis and the other two stops. Female speakers did not show a three-way contrast in any kinematic measure. They showed a two-way distinction between lenis and the other two stops in C-to-V deceleration duration (/k^*/=/k^h/</k/), and a two-way distinction between fortis and lenis stops in the opening movement duration. An overall comparison of VOT and articulatory analyses revealed that the lenis-aspirated kinematic distinction is diminishing, driven by female speakers, in line with the loss of the lenis-aspirated distinction in VOT that could influence supralaryngeal articulation.