Search

Research on deep learning-powered voice conversion (VC) in speech-to-speech scenarios are gaining increasing popularity. Although many of the works in the field of voice conversion share a common global pipeline, there is considerable diversity in the underlying structures, methods, and neural sub-blocks used across research efforts. Thus, obtaining a comprehensive understanding of the reasons behind the choice of the different methods included when training voice conversion models can be challenging, and the actual hurdles in the proposed solutions are often unclear. To shed light on these aspects, this paper presents a scoping review that explores the use of deep learning in speech analysis, synthesis, and disentangled speech representation learning within modern voice conversion systems. We screened 628 publications from more than 38 venues between 2017 and 2023, followed by an in-depth review of a final database of 130 eligible studies. Based on the review, we summarise themost frequently used approaches to voice conversion based on deep learning and highlight common pitfalls. We condense the knowledge gathered to identify main challenges, supply solutions grounded in the analysis and provide recommendations for future research directions. [ABSTRACT FROM AUTHOR]

The current methods for speaker diarization commonly employ standard global clustering techniques to distinguish speech segments of different speakers, without considering that voice of the same individual may exhibit various feature distribution under varying background noise conditions, which enlarges the intra-class distances and impacts clustering heavily. Motivated by that adjacent speech segments often share the same background noise, this paper proposed a novel temporalsegment-and-regroup clustering (TSARC) pipeline for speaker diarization to address above issues. Firstly, TSARC partitioned all speech segments into multiple independent intervals along their temporal continuity and conducted local clustering within each interval. Afterwards, it re-associated segments attributed to the same speaker across different intervals. Moreover, during the clustering process, the method actively employed neighborhood information from speech segments to calibrate their similarities. Through this way, TSARC reduced the likelihood of direct clustering under disparate noise conditions, effectively improving the accuracy of clustering. Experimental results on the public datasets AMI SDM and VoxConverse show that, compared with the baseline method, the proposed method has achieved relatively reductions in diarization error rate (DER) by 34% and 16% respectively, which proves the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

How do we perceive others based on their voices? This question has attracted research and media attention for decades, producing hundreds of studies showing that the voice is socially and biologically relevant, but these studies vary in methodology and ecological validity. Here we test whether vocalizers producing read versus free speech are judged similarly by listeners on ten biological and/or psychosocial traits. In perception experiments using speech from 208 men and women and ratings from 4,088 listeners, we show that listeners' assessments of vocalizer sex and age are highly accurate, regardless of speech type. Assessments of body size, femininity-masculinity and women's health also did not differ between free and read speech. In contrast, read speech elicited higher ratings of attractiveness, dominance and trustworthiness in both sexes and of health in males compared to free speech. Importantly, these differences were small, and we additionally show moderate to strong correlations between ratings of the same vocalizers based on their read and free speech for all ten traits, indicating that voice-based judgments are highly consistent within speakers, whether or not speech is spontaneous. Our results provide evidence that the human voice can communicate various biological and psychosocial traits via both read and free speech, with theoretical and practical implications. [ABSTRACT FROM AUTHOR]

Esophageal speech (ES) is a pathological voice that is often difficult to understand. Moreover, acquiring recordings of a patient's voice before a laryngectomy proves challenging, thereby complicating enhancing this kind of voice. That is why most supervised methods used to enhance ES are based on voice conversion, which uses healthy speaker targets, things that may not preserve the speaker's identity. Otherwise, unsupervised methods for ES are mostly based on traditional filters, which cannot alone beat this kind of noise, making the denoising process difficult. Also, these methods are known for producing musical artifacts. To address these issues, a self-supervised method based on the Only-Noisy-Training (ONT) model was applied, consisting of denoising a signal without needing a clean target. Four experiments were conducted using Deep Complex UNET (DCUNET) and Deep Complex UNET with Complex Two-Stage Transformer Module (DCUNET-cTSTM) for assessment. Both of these models are based on the ONT approach. Also, for comparison purposes and to calculate the evaluation metrics, the pre-trained VoiceFixer model was used to restore the clean wave files of esophageal speech. Even with the fact that ONT-based methods work better with noisy wave files, the results have proven that ES can be denoised without the need for clean targets, and hence, the speaker's identity is retained. [ABSTRACT FROM AUTHOR]

Purpose: This study explored how respiration, voice, and speech were affected following expiratory muscle strength training (EMST) and maintenance training in persons with Parkinson's disease (PD) or multiple sclerosis (MS). Method: Nine participants with PD and six with MS participated in a randomised study, where the effects of EMST, sham, and maintenance treatment were investigated. Outcome measures included maximum expiratory pressure (MEP); maximum phonation time (MPT); intelligibility; verbal diadochokinesis (DDK); speech rate; a self-report form on voice, speech, and communication; and open questions about how the participants experienced the intervention. Group comparisons were performed within and between groups. Result: The PD and the MS groups both improved significantly in MEP, and this improvement remained after 3 months of maintenance EMST. An improvement was also seen in DDK. Post-EMST, 33% of the PD group and 80% of the MS group reported a positive effect on communication. Conclusion: The results of this study support previous evidence that EMST has positive effects on expiratory pressure in persons with PD or MS, but its effect on voice and speech remains unclear. Since subjective reports of the intervention and effects on communication were predominantly positive, further research is needed on larger groups to explore appropriate outcome measures. [ABSTRACT FROM AUTHOR]

Many research articles have explored the impact of surgical interventions on voice and speech evaluations, but advances are limited by the lack of publicly accessible datasets. To address this, a comprehensive corpus of 107 Spanish Castilian speakers was recorded, including control speakers and patients who underwent upper airway surgeries such as Tonsillectomy, Functional Endoscopic Sinus Surgery, and Septoplasty. The dataset contains 3,800 audio files, averaging 35.51 ± 5.91 recordings per patient. This resource enables systematic investigation of the effects of upper respiratory tract surgery on voice and speech. Previous studies using this corpus have shown no relevant changes in key acoustic parameters for sustained vowel phonation, consistent with initial hypotheses. However, the analysis of speech recordings, particularly nasalised segments, remains open for further research. Additionally, this dataset facilitates the study of the impact of upper airway surgery on speaker recognition and identification methods, and testing of anti-spoofing methodologies for improved robustness. [ABSTRACT FROM AUTHOR]

Geminates (such as the double /k/ in Polish lekki “light”) form a group of consonants that are mainly characterized by longer durations than the corresponding singletons. Most of the research has concentrated on durational and spectral properties of geminates in contrast to singletons. Much less attention has been paid to the realization of the voicing contrast in geminates and whether it is differently implemented than in singletons. In the current study, we contribute to this research with the data from Polish stop geminates. To this end, a total of 49 native speakers of Polish produced all stop geminates and corresponding singletons in wordforms of the same phonological make-up. The measurements included closure duration, voicing ratio, duration, and mean intensity of the release burst. The results showed that the voicing ratio was 0.69, classifying Polish stop geminates as mildly devoiced. There was a significant speaker-dependent variability in that some speakers devoiced all geminates, while others either partially devoiced or never devoiced. The analysis of interactions between geminates and singletons revealed that geminates cancelled voicing cues observed in singletons such as longer durations and lower intensity of the release burst. We discuss the current results in terms of voicing implementation in Polish and in relation to other geminating languages. [ABSTRACT FROM AUTHOR]

Voice interventions for transgender individuals can address several speech and voice parameters – fundamental frequency, resonance, intonation, rhythm, and intensity. In this study, we focus on fundamental voice frequency and build on existing research to test one technique that has been shown in a preliminary study to effectively adjust fundamental voice frequency in line with clients' goals. More specifically, we employed an adaptation of melodic intonation therapy (MIT) to assess whether it can raise trans women's average fundamental frequency to a significant degree. Eleven trans women participated in two one-to-one therapy sessions, four weeks apart. Results pointed to a statistically significant rise in both their singing and speaking fundamental frequencies following the adapted MIT therapy sessions. Participants were also successful in imitating upward fundamental frequency contours when singing and in producing them independently in the speech modality. Longitudinal studies are warranted to determine whether the observed positive results translate into long-term benefits. [ABSTRACT FROM AUTHOR]

Background and Aim: Hearing-impaired individuals have difficulty comprehending and producing speech sounds. Cochlear implantation is used to augment hearing. The present study aims to compare the production of fricatives /s/ and /ʃ/ and affricate /ʧ/ by Persianspeaking Cochlear-Implanted (CI) and Normal-Hearing (NH) children Methods: Fifteen Persian-speaking NH children and 15 Persian-speaking CI children, matched for age, gender, and general health conditions, were included in the study. The stimuli included two voiceless Persian fricatives /s/ and /ʃ/ and one voiceless Persian affricate /ʧ/ along with the open front vowel /æ/ in three Consonant-Vowel (CV), Consonant-VowelConsonant (CVC), and Vowel-Consonant (VC) contexts (/sæ/, /æsæ/, /æs/, /ʃæ/, /æʃæ/, /æʃ/, /ʧæ/, /æʧæ/, /æʧ/). After recording all utterances, Praat software was used to measure the friction duration, rise time, and spectral peak of the consonants Results: The CI children could not distinguish between /ʃ/ and /ʧ/ and produced affricate /ʧ/ as an allophone of /ʃ/ (p=0.01). Moreover, distinguishing between two fricatives /s/ and /ʃ/ was difficult for both groups. While NH children slightly treated these two sounds differently, the CI group produced fricative /s/ as an allophone of /ʃ/ (p=0.02). The rise time of /ʃ/ was longer in the NH children, except for /ʧæ/, where the CI children had a longer rise time. Conclusion: The speech of CI children is different in producing /s/, /ʃ/, and /ʧ/ from their NH peers. The results can help speech therapists, clinical linguists, and application designers focus on speech sounds that are challenging for CI children to produce. [ABSTRACT FROM AUTHOR]

Introduction: Smart devices are widely available and capable of quickly recording and uploading speech segments for health-related analysis. The switch from laboratory recordings with professional-grade microphone setups to remote, smart device-based recordings offers immense potential for the scalability of voice assessment. Yet, a growing body of literature points to a wide heterogeneity among acoustic metrics for their robustness to variation in recording devices. The addition of consumer-grade plug-and-play microphones has been proposed as a possible solution. The aim of our study was to assess if the addition of consumer-grade plug-and-play microphones increases the acoustic measurement agreement between ultra-portable devices and a reference microphone. Methods: Speech was simultaneously recorded by a reference high-quality microphone commonly used in research and by two configurations with plug-and-play microphones. Twelve speech-acoustic features were calculated using recordings from each microphone to determine the agreement intervals in measurements between microphones. Agreement intervals were then compared to expected deviations in speech in various neurological conditions. Each microphone's response to speech and to silence was characterized through acoustic analysis to explore possible reasons for differences in acoustic measurements between microphones. The statistical differentiation of two groups, neurotypical and people with multiple sclerosis, using metrics from each tested microphone was compared to that of the reference microphone. Results: The two consumer-grade plug-and-play microphones favored high frequencies (mean center of gravity difference ≥ +175.3 Hz) and recorded more noise (mean difference in signal to noise ≤ −4.2 dB) when compared to the reference microphone. Between consumer-grade microphones, differences in relative noise were closely related to distance between the microphone and the speaker's mouth. Agreement intervals between the reference and consumer-grade microphones remained under disease-expected deviations only for fundamental frequency (f0, agreement interval ≤0.06 Hz), f0 instability (f0 CoV, agreement interval ≤0.05%), and tracking of second formant movement (agreement interval ≤1.4 Hz/ms). Agreement between microphones was poor for other metrics, particularly for fine timing metrics (mean pause length and pause length variability for various tasks). The statistical difference between the two groups of speakers was smaller with the plug-and-play than with the reference microphone. Conclusion: Measurement of f0 and F2 slope was robust to variation in recording equipment, while other acoustic metrics were not. Thus, the tested plug-and-play microphones should not be used interchangeably with professional-grade microphones for speech analysis. Plug-and-play microphones may assist in equipment standardization within speech studies, including remote or self-recording, possibly with small loss in accuracy and statistical power as observed in the current study. [ABSTRACT FROM AUTHOR]

Background: Maximum phonation time (MPT) is used to assess speech and other oral rehabilitation‐related issues. Various factors contribute to MPT decline in older individuals. Although the impact of physical frailty on MPT has been suggested, this has not been conclusively determined. Objective: To examine the relationship between MPT and physical frailty in community‐dwelling individuals aged ≥60 years who were independently mobile. MPT‐associated factors were investigated. Methods: This cross‐sectional study analysed the clinical data of 122 patients (age [interquartile range]: 80.0 [74.0–83.0] years) without dementia who visited a neurology department between 1 February 2021 and 31 January 2023. Investigated factors included age, sex, weight, height, body mass index, smoking history, grip strength, functional independence measure, vital capacity, oral diadochokinesis, MPT and the Japanese Cardiovascular Health Study score. Physical frailty was assessed based on the total score from five items (weight loss, weakness, exhaustion, slowness and low physical activity). The relationship between MPT and physical frailty was examined using Spearman's rank correlation coefficient and hierarchical multiple regression analysis. Results: The MPT was negatively correlated with age (r = −0.347, p <.01) and physical frailty (r = −0.681, p <.01) and positively correlated with vital capacity (r = 0.474, p <.01) and height (r = 0.248, p <.01). The hierarchical multiple regression analysis, conducted with MPT as the dependent variable, demonstrated that physical frailty (β = −.59, 95% confidence interval: −0.74 to 0.43, p <.001) had a strong influence on MPT. Conclusion: In older individuals, MPT is associated with physical frailty. When assessing MPT in clinical settings, it is advisable to perform a concurrent assessment of physical frailty. [ABSTRACT FROM AUTHOR]

This paper investigates the role of English in voice assistant (Siri, Alexa, Google Assistant) use from the perspective of language ideology. Major commercial companies in the voice assistant market use English as a training language for their speech technologies and offer the most optimised support for standardised varieties of English. This affects the experiences with voice assistants of speakers of non-European languages, i.e., one of the non-target audiences. Drawing on qualitative interview data from Turkish-speaking users who migrated to Germany, the present study reveals that the participants iconize English as the "standard" language in digital contexts, constructing it as the "original" language of speaking computers. By conducting an inductive analysis, the article demonstrates that not only the lack of technological support, but also specific discourses about Artificial Intelligence, impact perceptions of English. These developments have implications for our understandings of prestige and digital literacy in human-machine interactions. [ABSTRACT FROM AUTHOR]

Communication is often impaired in individuals with Parkinson's disease (PD), typically secondary to sensorimotor deficits impacting voice and speech. Language may also be diminished in PD, particularly for production and comprehension of verbs. Evidence exists that verb processing is influenced by motor system modulation suggesting that verb deficits in PD are underpinned by similarities in the neural representations of actions that span motor and semantic systems. Conversely, subtle differences in cognition in PD may explain difficulty in processing of complex syntactic forms, which increases cognitive demand and is linked to verb use. Here we investigated whether optimizing motor system support for vocal function (improving loudness) affects change in lexical semantic, syntactic, or informativeness aspects of spoken discourse. Picture description narratives were compared for 20 Control participants and 39 with PD, 19 of whom underwent Lee Silverman Voice Treatment (LSVT LOUD®). Treated PD narratives were also contrasted with those of untreated PD and Control participants at Baseline and after treatment. Controls differed significantly from the 39 PD participants for verbs per utterance, but this difference was largely driven by untreated PD participants who produced few utterances but with verbs, inflating their verbs per utterance. Given intervention, there was a significant increase in vocal loudness but no significant changes in language performance. These data do not support the hypothesis that targeting this speech motor system results in improved language production. Instead, the data provide evidence of considerable variability in measures of language production across groups, particularly in verbs per utterance. [ABSTRACT FROM AUTHOR]

Purpose: The purpose of this study is to examine whether phonetic information functions and how phonetic information affects voice identity processing in blind people. Method: To address the first inquiry, 25 normal sighted participants and 30 blind participants discriminated voice identity, when listening forward speech and backward speech from their own native language and another unfamiliar language. To address the second inquiry, combining articulatory suppression paradigm, 26 normal sighted participants and 26 blind participants discriminated voice identity, when listening forward speech from their own native language and another unfamiliar language. Results: In Experiment 1, not only in the voice identity discrimination task with forward speech, but also in the discrimination task with backward speech, both the sighted and blind groups showed the superiority of the native language. This finding supports the view that backward speech still retains some phonetic information, and indicates that phonetic information can affect voice identity processing in sighted and blind people. In addition, only the superiority of the native language of sighted people was regulated by the speech manner, which is related to articulatory rehearsal. In Experiment 2, only the superiority of the native language of sighted people was regulated by articulatory suppression. This indicates that phonetic information may act in different ways on voice identity processing in sighted and blind people. Conclusion: The heightened dependence on voice source information in blind people appears not to undermine the function of phonetic information, but it appears to change the functional mechanism of phonetic information. These findings suggest that the present phonetic familiarity model needs to be improved with respect to the mechanism of phonetic information. [ABSTRACT FROM AUTHOR]

This study investigated a sound change in progress by which the Central Bavarian dialect feature of complementary length between consonant and the preceding vowel is giving way to the unrestricted combination possibility of long (Vː) and short (V) vowels with following longer fortis (Cː) and shorter lenis (C) stops, respectively. This 2 × 2 system is also found in the standard variety of German. While previous studies have regarded any such findings of convergence toward Standard German as being a result of language contact, the present study specifically tested the possibility of fast-speech-induced hypoarticulation being a system-internal driver of this change. The focus of this study was on acoustic cues to the postvocalic stop. Following the apparent-time paradigm, acoustic analyses of 10 younger and 10 older dialect speakers revealed that (1) younger dialect speakers produced both VC and VːCː (both formerly illegal in the dialect), but (2) older dialect speakers produced only VːCː sequences with duration patterns similar to those of the control group of 10 Standard German speakers. Analyses of various dependent variables further showed (3) the (apparently) delayed emergence of aspiration as an additional cue to the fortis–lenis contrast in Western Central Bavarian particularly in younger dialect speakers, (4) no considerable effect of speech rate on the dispersion of and overlap between any of the four vowel-plus-stop combinations, and (5) the irregular spread of this change that appears to be gradual. As such, the findings support a model of linguistic change that also accounts for gradual changes in dialect borrowing. [ABSTRACT FROM AUTHOR]

Background noises are usually treated as redundant or even harmful to voice conversion. Therefore, when converting noisy speech, a pretrained module of speech separation is usually deployed to estimate clean speech prior to the conversion. However, this can lead to speech distortion due to the mismatch between the separation module and the conversion one. In this paper, a noise-robust voice conversion model is proposed, where a user can choose to retain or to remove the background sounds freely. Firstly, a speech separation module with a dual-decoder structure is proposed, where two decoders decode the denoised speech and the background sounds, respectively. A bridge module is used to capture the interactions between the denoised speech and the background sounds in parallel layers through information exchanging. Subsequently, a voice conversion module with multiple encoders to convert the estimated clean speech from the speech separation model. Finally, the speech separation and voice conversion module are jointly trained using a loss function combining cycle loss and mutual information loss, aiming to improve the decoupling efficacy among speech contents, pitch, and speaker identity. Experimental results show that the proposed model obtains significant improvements in both subjective and objective evaluation metrics compared with the existing baselines. The speech naturalness and speaker similarity of the converted speech are 3.47 and 3.43, respectively. [ABSTRACT FROM AUTHOR]

Voice conversion is the task of changing the speaker characteristics of input speech while preserving its linguistic content. It can be used in various areas, such as entertainment, medicine, and education. The quality of the converted speech is crucial for voice conversion algorithms to be useful in these various applications. Deep learning-based voice conversion algorithms, which have been showing promising results recently, generally consist of three modules: a feature extractor, feature converter, and vocoder. The feature extractor accepts the waveform as the input and extracts speech feature vectors for further processing. These speech feature vectors are later synthesized back into waveforms by the vocoder. The feature converter module performs the actual voice conversion; therefore, many previous studies separately focused on improving this module. These works combined the separately trained vocoder to synthesize the final waveform. Since the feature converter and the vocoder are trained independently, the output of the converter may not be compatible with the input of the vocoder, which causes performance degradation. Furthermore, most voice conversion algorithms utilize mel-spectrogram-based speech feature vectors without modification. These feature vectors have performed well in a variety of speech-processing areas but could be further optimized for voice conversion tasks. To address these problems, we propose a novel wave-to-wave (wav2wav) voice conversion method that integrates the feature extractor, the feature converter, and the vocoder into a single module and trains the system in an end-to-end manner. We evaluated the efficiency of the proposed method using the VCC2018 dataset. [ABSTRACT FROM AUTHOR]

Background: Deep brain stimulation (DBS) reliably ameliorates cardinal motor symptoms in Parkinson's disease (PD) and essential tremor (ET). However, the effects of DBS on speech, voice and language have been inconsistent and have not been examined comprehensively in a single study. Objective: We conducted a systematic analysis of literature by reviewing studies that examined the effects of DBS on speech, voice and language in PD and ET. Methods: A total of 675 publications were retrieved from PubMed, Embase, CINHAL, Web of Science, Cochrane Library and Scopus databases. Based on our selection criteria, 90 papers were included in our analysis. The selected publications were categorized into four subcategories: Fluency, Word production, Articulation and phonology and Voice quality. Results: The results suggested a long-term decline in verbal fluency, with more studies reporting deficits in phonemic fluency than semantic fluency following DBS. Additionally, high frequency stimulation, left-sided and bilateral DBS were associated with worse verbal fluency outcomes. Naming improved in the short-term following DBS-ON compared to DBS-OFF, with no long-term differences between the two conditions. Bilateral and low-frequency DBS demonstrated a relative improvement for phonation and articulation. Nonetheless, long-term DBS exacerbated phonation and articulation deficits. The effect of DBS on voice was highly variable, with both improvements and deterioration in different measures of voice. Conclusion: This was the first study that aimed to combine the outcome of speech, voice, and language following DBS in a single systematic review. The findings revealed a heterogeneous pattern of results for speech, voice, and language across DBS studies, and provided directions for future studies. [ABSTRACT FROM AUTHOR]

Speech recognition becomes increasingly important in the modern society, especially for human–machine interactions, but its deployment is still severely thwarted by the struggle of machines to recognize voiced commands in challenging real‐life settings: oftentimes, ambient noise drowns the acoustic sound signals, and walls, face masks or other obstacles hide the mouth motion from optical sensors. To address these formidable challenges, an experimental prototype of a microwave speech recognizer empowered by programmable metasurface is presented here that can remotely recognize human voice commands and speaker identities even in noisy environments and if the speaker's mouth is hidden behind a wall or face mask. The programmable metasurface is the pivotal hardware ingredient of the system because its large aperture and huge number of degrees of freedom allows the system to perform a complex sequence of sensing tasks, orchestrated by artificial‐intelligence tools. Relying solely on microwave data, the system avoids visual privacy infringements. The developed microwave speech recognizer can enable privacy‐respecting voice‐commanded human–machine interactions is experimentally demonstrated in many important but to‐date inaccessible application scenarios. The presented strategy will unlock new possibilities and have expectations for future smart homes, ambient‐assisted health monitoring, as well as intelligent surveillance and security. [ABSTRACT FROM AUTHOR]

Speech is a crucial source of personal information, and the risk of attackers using such information increases day by day. Speaker privacy protection is crucial, and various approaches have been proposed to hide the speaker's identity. One approach is voice anonymization, which aims to safeguard speaker identity while maintaining speech content through techniques such as voice conversion or spectral feature alteration. The significance of voice anonymization has grown due to the necessity to protect personal information in applications such as voice assistants, authentication, and customer support. Building upon the S3PRL-VC toolkit and on pre-trained speech and speaker representation models, this paper introduces a feature disentanglement approach to improve the de-identification performance of the state-of-the-art anonymization approaches based on voice conversion. The proposed approach achieves state-of-the-art speaker de-identification and causes minimal impact on the intelligibility of the signal after conversion. [ABSTRACT FROM AUTHOR]

Current methods of eliciting running speech for subjective rating, or perceptual analysis, in the assessment of pediatric voice disorders are not standardized. Whilst different assessment tools have different requirements, a commonality is the requirement to judge the perceptual characteristics of the individual's everyday speaking voice. However, it is unclear whether current practices yield ecologically valid running speech samples. The aim of this study was to analyse the length and characteristics of conversational responses, to stimuli that were designed to elicit running speech samples from pediatric clients. Twenty conversations, conducted as part of the evaluation of voice in a pediatric population, were analysed. Length of responses, number of responses and question types were recorded. The median maximum utterance length was 15.9 s; seven participants presented with a maximum of less than 10 s. Response length was significantly associated with question type, F = 10.68, p<.001. The most frequent number of responses produced was 11 (range = 17, IQR = 5, 11). There was a moderate correlation between response length and number of responses, r(18)=.53, p=.02. These data demonstrated that informal conversational methods did not consistently yield running speech samples of greater than 10 s in length, the minimum sample length recommended by the CAPE-V protocol. There was considerable variability in the characteristics of the responses produced by participants. Using a task such as a narrative re-tell might allow for better standardization of responses, including elicitation of vocal behaviors of interest, as well as yield a longer sample. [ABSTRACT FROM AUTHOR]

In laryngeal realism (LR), laryngeal specification of stops is explained by direct maps of cues (e.g. VOT) onto privative phonological laryngeal features [voice] or [spread glottis]. Phonetic realization of the segments and speakers' 'control' (e.g. the degree of intervocalic voicing and speech rate manipulation effects on VOT duration) are used as diagnostics of phonological specification. Similar to some Arabic vernacular dialects (e.g. Qatari Arabic), Khuzestani Arabic in Iran presents a case where three voiced stops /b d ɡ/ are in contrast with voiceless stops /p t k/, but two voiceless guttural plosives /tˁ q/ have no voiced homorganic counterparts. In this paper we examine the phonetic realization of voicing in these stops at word-initial and intervocalic position, as well as the effects of speech rate manipulation on VOT and closure voicing. The data came from 12 native speakers recorded in Khorramshahr, Iran. Our findings suggest an over-specified voicing system in this Arabic variety. We found that voiced /b d ɡ/ were produced with voicing lead in initial position and complete closure voicing word medially, voiceless /p t k/ had long lag VOT, while guttural /tˁ q/ had short lag VOT. Speech rate manipulation revealed that only duration of (pre)voicing and duration of aspiration increased in slower speech. Also, f0, F1, and F2 were measured at vowel onset to evaluate the glottal state in production of stops. The results support the predictions of LR that voiced stops are specified by [voice], voiceless stops are [sg], while gutturals lack underlying specification for voice. [ABSTRACT FROM AUTHOR]

Purpose: Physiological concepts can be effectively taught in speech science. These will serve as the foundation for understanding the physiological changes associated with neurogenic and voice disorders. In this tutorial, relevant concepts are highlighted across the respiratory, phonatory, velopharyngeal, and articulatory systems. Student activities and illustrative videos are presented. The culminating activity is based on a speech sample of an individual with moderate flaccid--spastic dysarthria. Students relate the physiological changes associated with dysarthria to the acoustic outcomes. Conclusions: Although traditionally speech science focuses on acoustics, this tutorial provides guidance on shaping the class to highlight the physiology underlying the acoustics. In this manner, students more easily see the clinical relevance. [ABSTRACT FROM AUTHOR]

Nowadays, fine-tuning has emerged as a powerful technique in machine learning, enabling models to adapt to a specific domain by leveraging pre-trained knowledge. One such application domain is automatic speech recognition (ASR), where fine-tuning plays a crucial role in addressing data scarcity, especially for languages with limited resources. In this study, we applied fine-tuning in the context of atypical speech recognition, focusing on Italian speakers with speech impairments, e.g., dysarthria. Our objective was to build a speaker-dependent voice user interface (VUI) tailored to their unique needs. To achieve this, we harnessed a pre-trained OpenAI's Whisper model, which has been exposed to vast amounts of general speech data. However, to adapt it specifically for disordered speech, we fine-tuned it using our private corpus including 65 K voice recordings contributed by 208 speech-impaired individuals globally. We exploited three variants of the Whisper model (small, base, tiny), and by evaluating their relative performance, we aimed to identify the most accurate configuration for handling disordered speech patterns. Furthermore, our study dealt with the local deployment of the trained models on edge computing nodes, with the aim to realize custom VUIs for persons with impaired speech. [ABSTRACT FROM AUTHOR]

Introduction: Bilinguals constitute a significant portion of speech-language pathologists' (SLPs) caseloads. Insight into the cross-linguistic effect on voice is needed to guide SLPs to make linguistically appropriate observations when working with heterogenous populations. Method: Nineteen female English-Northern Sotho bilinguals performed three speech tasks (reading, picture description, and monologue) in each language. Acoustic analysis of mean fundamental frequency (f0), intensity, and articulation rate was conducted with Praat. A panel of blinded listeners reached consensus after independently reviewing the recordings during perceptual analysis of voice quality, resonance, and glottal attack. Results: The following statistically significant differences were found across and within the languages: The mean f0 was 204.61 Hz in the Northern Sotho picture description yet 196.50 Hz in the English picture description. The mean intensity of reading in Northern Sotho was 66.38 dB whereas the mean intensity of reading in English was 65.09 dB. Articulation rate was 3.78 syllables/s in English passage reading and 3.41 syllables/s in Northern Sotho passage reading. Within English, passage reading elicited a significantly quicker articulation rate than the picture description (3.34 syllables/s) and monologue (3.46 syllables/s). Within Northern Sotho, mean f0 was 203.83 Hz in passage reading yet 191.11 Hz in the monologue. Perceptual voice quality, glottal attack, and resonance were comparable across languages. Conclusion: Relationships between languages spoken, task performance, and vocal characteristics were observed in English-Northern Sotho bilingual females. SLPs must consider the interaction of language, task performance, and vocal characteristics when working with bilingual clients. [ABSTRACT FROM AUTHOR]

The use of voice recordings in both research and industry practice has increased dramatically in recent years—from diagnosing a COVID-19 infection based on patients' self-recorded voice samples to predicting customer emotions during a service center call. Crowdsourced audio data collection in participants' natural environment using their own recording device has opened up new avenues for researchers and practitioners to conduct research at scale across a broad range of disciplines. The current research examines whether fundamental properties of the human voice are reliably and validly captured through common consumer-grade audio-recording devices in current medical, behavioral science, business, and computer science research. Specifically, this work provides evidence from a tightly controlled laboratory experiment analyzing 1800 voice samples and subsequent simulations that recording devices with high proximity to a speaker (such as a headset or a lavalier microphone) lead to inflated measures of amplitude compared to a benchmark studio-quality microphone while recording devices with lower proximity to a speaker (such as a laptop or a smartphone in front of the speaker) systematically reduce measures of amplitude and can lead to biased measures of the speaker's true fundamental frequency. We further demonstrate through simulation studies that these differences can lead to biased and ultimately invalid conclusions in, for example, an emotion detection task. Finally, we outline a set of recording guidelines to ensure reliable and valid voice recordings and offer initial evidence for a machine-learning approach to bias correction in the case of distorted speech signals. [ABSTRACT FROM AUTHOR]

Second-language learners often encounter communication challenges due to a foreign accent (FA) in their speech, influenced by their native language (L1). This FA can affect rhythm, intonation, stress, and the segmental domain, which consists of individual language sounds. This study looks into the segmental FA aspect, exploring listeners' perceptions when Spanish interacts with English. Utilizing the SIAEW corpus, which replaces segments of English words with anticipated Spanish-accented realizations, we assess the ability of non-native listeners to discriminate degrees of accent across male and female voices. This research aims to determine the impact of voice consistency on detecting accentedness variations, studying participants from Japanese and Spanish. Results show that, while listeners are generally able to discriminate degrees of foreign accent across speakers, some segmental transformations convey a more clear distinction depending on the phonological representations of the native and accented realisations on the listener's system. Another finding is that listeners tend to better discriminate degrees of accent when words are more native-like sounding. [ABSTRACT FROM AUTHOR]

Objective: Patients lose their voice after laryngectomy for laryngeal cancer or aspiration prevention surgery for severe dysphagia. To assist such patients, we developed and verified the utility of a novel vocalization method using a device termed the voice retriever (VR), in which the sound source is placed in the mouth. Methods: We investigated the effectiveness of the VR in patients. The VR consists of a mouthpiece with a built‐in speaker and a dedicated application that serves as the sound source. We compared the speech intelligibility and naturalness in normal participants using VR and an electrolarynx (EL) for the first time as well as the voice‐related quality of life (V‐RQOL) in patients with dysphonia before and after using the VR. Results: The VR produced significantly higher 100‐syllable test scores as well as fluency, amount of additional noise, intonation, intelligibility and overall long reading test ratings in the first‐time VR and EL users. Furthermore, the VR use significantly improved the V‐RQOL of participants with dysphonia. Conclusion: Compared to EL, VR allows more effective speech improvement in participants without experience using an alternative vocalization method and improves the V‐RQOL in patients with dysphonia. Level of Evidence: Step 4. [ABSTRACT FROM AUTHOR]

Interpreters unanimously read ἐνευλογηθήσονται in Gen 12:3b LXX as a passive. Good evidence, however, exists to challenge and problematize this conclusion. Recent linguistic studies on the ancient Greek middle voice reveal that aorist and future -θη- forms express a semantically middle domain. When we reexamine the word ἐνευλογέομαι within this light, a better option emerges for seeing its -θη- forms as manifestations of speech actions within this middle domain. In their own unique ways, the LXX as well as Philo, Paul, and Acts further corroborate this alternative. As a result, we may read ἐνευλογηθήσονται in Gen 12:3b LXX as a speech action middle: "to pronounce blessings." The proposed reading promotes a better understanding of Abraham within Genesis LXX. Rather than a means to an end, Abraham remains at the center of God's blessing as the earth's families cry out: "God make me like Abraham!" [ABSTRACT FROM AUTHOR]

Introduction The voice and hearing can be affected to different degrees by aging, which can cause communication difficulties for elderly people. Vocal production requires effective temporal auditory processing at central levels within the nervous system, which can be compromised by the aging process. Objective To analyze the correlation between voice and temporal auditory processing in older adults. Materials and Methods A total of 40 elderly people aged 60 years or older were subdivided into 2 groups according to the presence or absence of vocal symptoms measured by the Voice Symptom Scale. All of the participants were submitted to auditory temporal tests, vocal self-assessment, and acoustic and perceptual auditory analyses of voice. Results Most of the subjects assessed had decreased voice intensity and normal variability in terms of vocal quality. The performance was normal in the Pitch Pattern Sequence test and altered in the Random Gap Detection test. In the Masking Period Pattern test, the detection thresholds for the target signal were increased in the presence of masking in different temporal target signal positions. Only pitch differed between the two groups. There were differences between the genders regarding frequency, shimmer, the overall severity of the alteration, and roughness. There was a correlation regarding temporal resolution ability and the overall severity of the alteration and roughness of the voice. Conclusion There is a central auditory impairment in temporal resolution which is correlated with vocal alterations in the elderly. [ABSTRACT FROM AUTHOR]

Purpose: Here, we examine the possibility that memory consolidation during a period of postpractice rest or nocturnal sleep can bolster speech motor learning in the absence of additional practice or effort. Method: Using web-administered experiments, 74 typical, American English talkers trained in a nonnative vowel contrast then had a 12-hr delay with (SLEEP) or without nocturnal sleep (REST) or proceeded immediately (IMMEDIATE) to a posttraining production assessment. For ecological validity, 51 native Danish talkers perceptually identified the American English talkers’ productions. Results: We observed that practice resulted in productions that were more acoustically similar to the Danish target. In addition, we found that rest in the absence of further practice reduced the token-to-token variability of the productions. Last, for vowels produced immediately following training, listeners more accurately identified vowels in the trained context, whereas in the untrained context, listener accuracy improved only for vowels produced by talkers who slept. Conclusions: A single session of speech motor training promotes observable change to speech production behavior. Specifically, practice facilitates acoustic similarity to the target. Moreover, although a 12-hr postpractice period of rest appears to promote productions that are less variable, only the productions of those who slept are perceived as more accurate by listeners. This may point to sleep’s role in contextualizing the acoustic goal of the production to the learner’s own vocal tract and its role as a protective mechanism during learning. These results are unaccounted for under existing models and offer potential for future educational and clinical applications to maximize speech motor learning. [ABSTRACT FROM AUTHOR]

Purpose: Behavioral assays of feedforward and feedback auditory–motor control of voice and articulation frequently are used to make inferences about underlying neural mechanisms and to study speech development and disorders. However, no studies have examined the test–retest reliability of such measures, which is critical for rigorous study of auditory–motor control. Thus, the purpose of the present study was to assess the reliability of assays of feedforward and feedback control in voice versus articulation domains. Method: Twenty-eight participants (14 cisgender women, 12 cisgender men, one transgender man, one transmasculine/nonbinary) who denied any history of speech, hearing, or neurological impairment were measured for responses to predictable versus unexpected auditory feedback perturbations of vocal (fundamental frequency, fo) and articulatory (first formant, F1) acoustic parameters twice, with 3–6 weeks between sessions. Reliability was measured with intraclass correlations. Results: Opposite patterns of reliability were observed for fo and F1; fo reflexive responses showed good reliability and fo adaptive responses showed poor reliability, whereas F1 reflexive responses showed poor reliability and F1 adaptive responses showed moderate reliability. However, a criterion-referenced categorical measurement of fo adaptive responses as typical versus atypical showed substantial test–retest agreement. Conclusions: Individual responses to some behavioral assays of auditory–motor control of speech should be interpreted with caution, which has implications for several fields of research. Additional research is needed to establish reliable criterion-referenced measures of F1 adaptive responses as well as fo and F1 reflexive responses. Furthermore, the opposite patterns of test–retest reliability observed for voice versus articulation add to growing evidence for differences in underlying neural control mechanisms. [ABSTRACT FROM AUTHOR]

Background: Research has shown that potential perpetrators and individuals high in psychopathic traits tend to body language cues to target a potential new victim. However, whether targeting occurs also by tending to vocal cues has not been examined. Thus, the role of voice in interpersonal violence merits investigation. Objective: In two studies, we examined whether perpetrators could differentiate female speakers with and without sexual and physical assault histories (presented as rating the degree of ‘vulnerability’ to victimization). Methods: Two samples of male listeners (sample one N = 105, sample two, N = 109) participated. Each sample rated 18 voices (9 survivors and 9 controls). Listener sample one heard spontaneous speech, and listener sample two heard the second sentence of a standardized passage. Listeners’ self-reported psychopathic traits and history of previous perpetration were measured. Results: Across both samples, history of perpetration (but not psychopathy) predicted accuracy in distinguishing survivors of assault. Conclusions: These findings highlight the potential role of voice in prevention and intervention. Gaining a further understanding of what voice cues are associated with accuracy in discerning survivors can also help us understand whether or not specialized voice training could have a role in self-defense practices. [ABSTRACT FROM AUTHOR]

This study examines whether Namibian English spoken by L1 Khoekhoegowab speakers has word-initial stop voicing contrasts, and how they might be realized in freely produced speech. The data consists of English interviews collected from nine speakers of Namibian English. For analysis, a total of 365 words beginning with a stop-vowel sequence were extracted from the interviews, and analyzed for voice onset time (VOT). To examine the realization of voicing, the extracted words are divided into voiced and voiceless categories based on their phonological voicing in other varieties of English and their VOTs are compared statistically. The VOTs of voiceless stops are significantly longer than those of voiced stops. The results suggest a short-lag vs. aspiration contrast in word-initial stops in Namibian English and provide new information about the phonetic features of Namibian English consonants. [ABSTRACT FROM AUTHOR]

In this article, some theories about who could have been the first singer are considered. [ABSTRACT FROM AUTHOR]

Nonverbal acoustic parameters of the human voice provide cues to a vocaliser's sex, age, and body size that are relevant in human social and sexual communication, and also increasingly so for computer-based voice recognition and synthesis technologies. While studies have shown some capacity in human listeners to gauge these biological traits from unseen speakers, it remains unknown whether speech complexity improves accuracy. Here, in over 200 vocalisers and 1500 listeners of both sexes, we test whether voice-based assessments of sex, age, height and weight vary from isolated vowels and words, to sequences of vowels and words, to full sentences or paragraphs. We show that while listeners judge sex and especially age more accurately as speech complexity increases, accuracy remains high across speech types, even for a single vowel sound. In contrast, the actual heights and weights of vocalisers explain comparatively less variance in listener's assessments of body size, which do not vary systematically by speech type. Our results thus show that while more complex speech can improve listeners' biological assessments, the gain is ecologically small, as listeners already show an impressive capacity to gauge speaker traits from extremely short bouts of standardised speech, likely owing to within-speaker stability in underlying nonverbal vocal parameters such as voice pitch. We discuss the methodological, technological, and social implications of these results. [ABSTRACT FROM AUTHOR]

This paper represents a thorough method for integrating emotions, texttospeech conversion, and state of the art voice cloning. The paper focuses on novel background noise adaptation, emotional voice synthesis, and multispeaker voice cloning for better speech synthesis. The synthesis of emotive voices, multi-speaker voice cloning, and creative methods for modifying background noise to improve speech synthesis quality are among the topics covered in this study. Additionally, the study explores the domain of emotional artificial intelligence by adding a variety of emotions to artificial voices, improving user engagement through sympathetic reactions. The study also looks at how background noise can be altered to change it from a disturbing to a silent, non-disruptive state. The texttospeech systems usability in noisy conditions is greatly enhanced by this improvement. By integrating these components, the project makes a substantial contribution to text to speech, emotional AI, and voice cloning, creating new avenues for human-computer connection. [ABSTRACT FROM AUTHOR]

This paper presents an acoustic description of the production of the plain uvular /q/ in the speech of eight speakers of South Bolivian Quechua. While this sound patterns phonologically as a stop, its primary realization is as a voiced continuant. Variation is documented with respect to segmental and prosodic position. Segmentally, a voiced continuant is the most common realization intervocalically and after a rhotic, while a voiceless continuant is comparatively more frequent after a voiceless sibilant, and voiced stops are most common after a nasal. In post-pausal position, voiced continuant productions are still attested and are particularly common for certain speakers, suggesting that this sound category has been reanalyzed as a continuant. For other speakers, voiceless stop productions are common or preferred in post-pausal position, reflecting a standard prosodically conditioned lenition pattern. Interestingly, voiced stops also show increased frequency in post-pausal position. The production of the plain uvular is analyzed in spontaneous speech collected in an interview format, as well as in scripted speech from a word list task. A second analysis compares the realization of /q/ to the other three stops /p t k/ in the language in spontaneous speech, and finds significantly more continuant productions for /q/. [ABSTRACT FROM AUTHOR]

The present study intends to dive into the world of technically appropriate human sound production. Whether it’s used in professional speech or singing, understanding the different elements that aid in the correct creation and emission of sound are prerequisites for all who intend to pursue a career that requires speaking or singing for an extended period of time. From the characteristics of sound, to the anatomy of how sound is created, what does technical breathing mean but also what a voice trainer needs to pay attention to in terms of anatomy but also psychology when working with students, the following article presents the steps and methods a vocal coach needs to take into account when taking on a new student. [ABSTRACT FROM AUTHOR]

Purpose: Individuals born with a cleft palate with or without a cleft lip (CP ± L) often experience functional, aesthetic, and psychosocial consequences well into adulthood. This study aimed to investigate outcomes of speech and health-related quality of life (HRQoL) in adults with a CP ± L who received interdisciplinary cleft care at the Ghent University Hospital using valid, reliable, and condition-specific instruments. Method: Thirteen Belgian Dutch-speaking participants with a CP ± L with a mean age of 25.4 years (SD = 5.1, range: 20-33 years) and an age- and gender-matched control group of 13 participants without a CP ± L with a mean age of 25.2 years (SD = 4.8, range: 20-32 years) were included in this study. Speech characteristics were evaluated perceptually and instrumentally. HRQoL was assessed through standardized patient-reported outcome measures. Outcomes were compared with those of the control group and to normative data where available. Results: Participants with a CP ± L in this sample demonstrated significantly lower speech acceptability (p < .001) and higher rates of hypernasality (p = .015) and nasal turbulence (p = .005) than the control group. They showed significantly higher satisfaction with appearance of the cleft scar compared with norms of adults with a CP ± L (p = .047). No other differences in speech characteristics, sociodemographics, or HRQoL were found between participants with and without a CP ± L. Conclusions: The reduced speech acceptability and the presence of resonance and nasal airflow disorders may indicate the need for standardized long-term outcome measurement and interdisciplinary follow-up for speech characteristics and velopharyngeal insufficiency in young and middle adulthood in future clinical practice. Additional research is necessary to further substantiate these findings and to determine predictors for these continuing complications in adults with a CP ± L. [ABSTRACT FROM AUTHOR]

We present an any-to-one voice conversion (VC) system, using an autoregressive model and LPCNet vocoder, aimed at enhancing the converted speech in terms of naturalness, intelligibility, and speaker similarity. As the name implies, non-parallel any-to-one voice conversion does not require paired source and target speeches and can be employed for arbitrary speech conversion tasks. Recent advancements in neural-based vocoders, such as WaveNet, have improved the efficiency of speech synthesis. However, in practice, we find that the trajectory of some generated waveforms is not consistently smooth, leading to occasional voice errors. To address this issue, we propose to use an autoregressive (AR) conversion model along with the high-fidelity LPCNet vocoder. This combination not only solves the problems of waveform fluidity but also produces more natural and clear speech, with the added capability of real-time speech generation. To precisely represent the linguistic content of a given utterance, we use speaker-independent PPG features (SI-PPG) computed from an automatic speech recognition (ASR) model trained on a multi-speaker corpus. Next, a conversion model maps the SI-PPG to the acoustic representations used as input features for the LPCNet. The proposed autoregressive structure enables our system to produce the following prediction step outputs from the acoustic features predicted in the previous step. We evaluate the effectiveness of our system by performing any-to-one conversion pairs between native English speakers. Experimental results show that the proposed method outperforms state-of-the-art systems, producing higher speech quality and greater speaker similarity. [ABSTRACT FROM AUTHOR]

This text studies the corporeality of attentive reading. It relies and builds upon philosopher Jean-Luc Nancy's suggestion that there is, each time, a recitative voice within the heart of our advancement through a textual body. This text examines the intriguing figure of recitative voice by paying attention to two bodily variations of reading: reading aloud and reading silently. Nancy's recitative voice, as a sonorous, resonant, oral, buccal and vocal notion, can help us in explicating how our bodies condition our experiences of reading, yet ultimately he remains rather implicit on how we advance through textual bodies. This short text argues that we can explicate the bodily weight implicit to reading, if we interpreted it as a recitative act. When we read aloud, our experience of a text consists of an advancement of our conjoined gaze and voice through the letters, words and sentences making up our text; when we read silently, our closed and silent mouth is still conjoined with our advancing gaze, advancing as desirous for speech and with words already grasped, even when no words are folded in (or with) exhaled warm air. [ABSTRACT FROM AUTHOR]

Human voice recognition over telephone channels typically yields lower accuracy when compared to audio recorded in a studio environment with higher quality. Here, we investigated the extent to which audio in video conferencing, subject to various lossy compression mechanisms, affects human voice recognition performance. Voice recognition performance was tested in an old–new recognition task under three audio conditions (telephone, Zoom, studio) across all matched (familiarization and test with same audio condition) and mismatched combinations (familiarization and test with different audio conditions). Participants were familiarized with female voices presented in either studio-quality (N = 22), Zoom-quality (N = 21), or telephone-quality (N = 20) stimuli. Subsequently, all listeners performed an identical voice recognition test containing a balanced stimulus set from all three conditions. Results revealed that voice recognition performance (dʹ) in Zoom audio was not significantly different to studio audio but both in Zoom and studio audio listeners performed significantly better compared to telephone audio. This suggests that signal processing of the speech codec used by Zoom provides equally relevant information in terms of voice recognition compared to studio audio. Interestingly, listeners familiarized with voices via Zoom audio showed a trend towards a better recognition performance in the test (p = 0.056) compared to listeners familiarized with studio audio. We discuss future directions according to which a possible advantage of Zoom audio for voice recognition might be related to some of the speech coding mechanisms used by Zoom. [ABSTRACT FROM AUTHOR]

Speech perception is challenging under adverse conditions. However, there is limited evidence regarding how multiple adverse conditions affect speech perception. The present study investigated two conditions that are frequently encountered in real-life communication: background noise and breathy vocal quality. The study first examined the effects of background noise and breathiness on speech perception as measured by intelligibility. Secondly, the study tested the hypothesis that both noise and breathiness affect listening effort, as indicated by linear and nonlinear changes in pupil dilation. Low-context sentences were resynthesized to create three levels of breathiness (original, mild-moderate, and severe). The sentences were presented in a fluctuating nonspeech noise with two signal-to-noise ratios (SNRs) of −5 dB (favorable) and −9 dB (adverse) SNR. Speech intelligibility and pupil dilation data were collected from young listeners with normal hearing thresholds. The results demonstrated that a breathy vocal quality presented in noise negatively affected speech intelligibility, with the degree of breathiness playing a critical role. Listening effort, as measured by the magnitude of pupil dilation, showed significant effects with both severe and mild-moderate breathy voices that were independent of noise level. The findings contributed to the literature by demonstrating the impact of vocal quality on the perception of speech in noise. They also highlighted the complex dynamics between overall task demand and processing resources in understanding the combined impact of multiple adverse conditions. [ABSTRACT FROM AUTHOR]

According to the Speech Learning Model (Flege 1995), successful L2 phonological acquisition is facilitated by the formation of new phonetic categories in the L2. However, category formation may be hindered by equivalence classification, wherein speakers perceptually merge L1 and L2 sounds. This study examines L1 Polish learners of English, including a phonetic parameter that has received minimal attention: affrication of /tr-dr/ clusters in English. Two groups of speakers, comprising B1 level and C2 level learners, produced word lists containing both initial /tr-dr/ clusters, as well as singleton voiced stops /b, d, g/. The results revealed an asymmetry: both groups failed to suppress pre-voicing in /b, d, g/, but were successful in producing affricated clusters. A new category has therefore been formed for the clusters, but not for the singleton stops. Phonological implications of this finding are discussed. [ABSTRACT FROM AUTHOR]

The ability to discriminate between talkers based on their fundamental (F0) and formant frequencies can facilitate speech comprehension in multi-talker environments. To date, voice discrimination (VD) of children and adults has only been tested in quiet conditions. This study examines the effect of speech-shaped noise on the use of F0 only, formants only, and the combined F0 + formant cues for VD. A total of 24 adults (18–35 years) and 16 children (7–10 years) underwent VD threshold assessments in quiet and noisy environments with the tested cues. Thresholds were obtained using a three-interval, three-alternative, two-down, one-up adaptive procedure. The results demonstrated that noise negatively impacted the utilization of formants for VD. Consequently, F0 became the lead cue for VD for the adults in noisy environments, whereas the formants were the more accessible cue for VD in quiet environments. For children, however, both cues were poorly utilized in noisy environments. The finding that robust cues such as formants are not readily available for VD in noisy conditions has significant clinical implications. Specifically, the reliance on F0 in noisy environments highlights the difficulties that children encounter in multi-talker environments due to their poor F0 discrimination and emphasizes the importance of maintaining F0 cues in speech-processing strategies tailored for hearing devices. [ABSTRACT FROM AUTHOR]