Voice Samples Research Articles

Voice analysis and deep learning for detecting mental disorders in pregnant women: a cross-sectional study

IntroductionPerinatal mental disorders are prevalent, affecting 10–20% of pregnant women, and can negatively impact both maternal and neonatal outcomes. Traditional screening tools, such as the Edinburgh Postnatal Depression Scale (EPDS), present limitations due to subjectivity and time constraints in clinical settings. Recent advances in voice analysis and machine learning have shown potential for providing more objective screening methods. This study aimed to develop a deep learning model that analyzes the voices of pregnant women to screen for mental disorders, thereby offering an alternative to the traditional tools.MethodsA cross-sectional study was conducted among 204 pregnant women, from whom voice samples were collected during their one-month postpartum checkup. The audio data were preprocessed into 5000 ms intervals, converted into mel-spectrograms, and augmented using TrivialAugment and context-rich minority oversampling. The EfficientFormer V2-L model, pretrained on ImageNet, was employed with transfer learning for classification. The hyperparameters were optimized using Optuna, and an ensemble learning approach was used for the final predictions. The model’s performance was compared to that of the EPDS in terms of sensitivity, specificity, and other diagnostic metrics.ResultsOf the 172 participants analyzed (149 without mental disorders and 23 with mental disorders), the voice-based model demonstrated a sensitivity of 1.00 and a recall of 0.82, outperforming the EPDS in these areas. However, the EPDS exhibited higher specificity (0.97) and precision (0.84). No significant difference was observed in the area under the receiver operating characteristic curve between the two methods (p = 0.759).DiscussionThe voice-based model showed higher sensitivity and recall, suggesting that it may be more effective in identifying at-risk individuals than the EPDS. Machine learning and voice analysis are promising objective screening methods for mental disorders during pregnancy, potentially improving early detection.ConclusionWe developed a lightweight machine learning model to analyze pregnant women's voices for screening various mental disorders, achieving high sensitivity and demonstrating the potential of voice analysis as an effective and objective tool in perinatal mental health care.

A Deep-Learning Model for Multi-class Audio Classification of Vocal Fold Pathologies in Office Stroboscopy.

To develop and validate a deep-learning classifier trained on voice data extracted from videolaryngostroboscopy recordings, differentiating between three different vocal fold (VF) states: healthy (HVF), unilateral paralysis (UVFP), and VF lesions, including benign and malignant pathologies. Patients with UVFP (n = 105), VF lesions (n = 63), and HVF (n = 41) were retrospectively identified. Voice samples were extracted from stroboscopic videos (Pentax Laryngeal Strobe Model 9400), including sustained /i/ phonation, pitch glide, and /i/ sniff task. Extracted audio files were converted into Mel-spectrograms. Voice samples were independently divided into training (80%), validation (10%), and test (10%) by patient. Pretrained ResNet18 models were trained to classify (1) HVF and pathological VF (lesions and UVFP), and (2) HVF, UVFP, and VF lesions. Both classifiers were further validated on an external dataset consisting of 12 UVFP, 13 VF lesions, and 15 HVF patients. Model performances were evaluated by accuracy and F1-score. When evaluated on a hold-out test set, the binary classifier demonstrated stronger performance compared to the multi-class classifier (accuracy 83% vs. 40%; F1-score 0.90 vs. 0.36). When evaluated on an external dataset, the binary classifier achieved an accuracy of 63% and F1-score of 0.48, compared to 35% and 0.25 for the multi-class classifier. Deep-learning classifiers differentiating HVF, UVFP, and VF lesions were developed using voice data from stroboscopic videos. Although healthy and pathological voice were differentiated with moderate accuracy, multi-class classification lowered model performance. The model performed poorly on an external dataset. Voice captured in stroboscopic videos may have limited diagnostic value, though further studies are needed. 4 Laryngoscope, 2025.

Severity Cut-Off Ranges for Auditory-Perceptual Evaluation Using Visual Analog Scales.

This retrospective, observational analytic, cross-sectional research study aimed to establish cut-off scores for categorizing the severity of voice qualities on a 100-unit visual analog scale (VAS). The study investigated the VAS cut-off ranges corresponding to auditory-perceptual ratings of normal, mild, moderate, and severe severities from the GRBAS scale for overall severity, breathiness, roughness, and strain. The study utilized voice samples and ratings from thePerceptual Voice Qualities Database(PVQD). The database contains 296 audio samples, including sustained vowels and sentences, rated by 19 experienced voice clinicians. The raters used a 100-point VAS and the GRBAS scale to evaluate voice qualities. Receiver operator curves (ROC) were calculated to determine cut-off scores for each severity category, balancing a 90% sensitivity with a 10% allowable false-positive categorization. Cut-off ranges were established for overall severity, breathiness, and roughness, demonstrating similar values across severities. The ranges were: Overall severity: normal (0-15), mild (16-39), moderate (40-69), severe (70-100). Breathiness: normal (0-14), mild (15-34), moderate (35-68), severe (69-100). Roughness: normal (0-13), mild (14-37), moderate (38-65), severe (66-100). For strain, only the normal range (0-15) could be determined with statistical significance. This study provided clinically relevant cut-off values for severity categories on a VAS, enhancing communication with patients and facilitating the use of both sensitive and specific categorical measures in voice evaluation. The findings contribute to a more standardized and objective assessment of voice quality.

Motivating tones to enhance education: The effects of vocal awareness on teachers' voices.

Effective classroom communication is key to shaping the learning environment and inspiring student engagement. And, it's not just what is said, but how it's said, that influences students. Yet, few (current or future) teachers receive education on vocal pedagogy. This study examined the impact of raising vocal awareness in teachers on their voice production through delivering a voice training program. Specifically, we explored how primary school teacher trainees produced motivational (either soft, warm, and encouraging, or harsh, pressuring, and controlling) and neutral communications before and after the delivery of a voice education program that concentrated on raising voice awareness, vocal anatomy, exercise techniques (e.g. breath control, voice modulation), and voice care. We hypothesised that trainees' voice production would change over the course of the program and lead to more 'prototypical' displays of motivational prosody (e.g. softly spoken encouraging intentions vs. harshly spoken controlling intentions). Results indicated a noticeable difference when communicating motivational intentions between pre- and post-training voice samples: post training, trainees spoke more slowly and with reduced vocal effort irrespective of motivational intention, suggesting that raising vocal awareness can alter classroom communications. The results underscore the importance of vocal awareness training to create a supportive and autonomy-enhancing learning environment.

STRAIGHTMORPH: A Voice Morphing Tool for Research in Voice Communication Sciences.

The purpose of this paper is to make easily available to the scientific community an efficient voice morphing tool called STRAIGHTMORPH and provide a short tutorial on its use with examples. STRAIGHTMORPH consists of a set of Matlab functions allowing the generation of high-quality, parametrically-controlled morphs of an arbitrary number of voice samples. A first step consists in extracting an 'mObject' for each voice sample, with accurate tracking of the fundamental frequency contour and manual definition of Time and Frequency anchors corresponding across samples to be morphed. The second step consists in parametrically combining the mObjects to generate novel synthetic stimuli, such as gender, identity or emotion continua, or random combinations. STRAIGHTMORPH constitutes a simple but efficient and versatile tool to generate high quality, parametrically controlled continua between voices - and beyond.

Early Detection of Parkinson’s Disease Using Machine Learning

Parkinson’s Disease (PD) is a chronic neurodegen- erative disorder that significantly impacts motor and non-motor functions, affecting millions of people globally. While motor symptoms such as tremors, rigidity, and bradykinesia are the most recognizable features, subtle non-motor signs, including vocal impairments, often emerge in the early stages of the disease. Early diagnosis of PD is crucial, as timely interventions can slow disease progression and improve patients’ quality of life. Traditional diagnostic methods, reliant on clinical observations and imaging, are often subjective, expensive, and inaccessible in resource-limited settings. This research introduces a real-time, non-invasive system for PD prediction using voice analysis. The proposed system leverages advanced machine learning techniques to identify vocal biomarkers indicative of Parkinson’s. Features such as Mel- frequency cepstral coefficients (MFCCs), jitter, shimmer, and harmonics-to-noise ratio (HNR) are extracted from recorded voice samples, providing a quantitative basis for early disease detection. By employing deep learning architectures like CNN- LSTM, the system achieves high accuracy in distinguishing Parkinsonian voices from healthy ones, even in challenging environments. Designed for real-time operation, the system integrates voice input, feature extraction, and predictive analysis into a seamless pipeline with minimal latency. The implementation emphasizes scalability, allowing the system to handle diverse linguistic and demographic variations. Evaluation on benchmark datasets, such as the Parkinson’s Telemonitoring dataset, demonstrates the sys- tem’s robustness, achieving significant improvements in accuracy and processing speed compared to existing methods. The system’s user-friendly interface ensures accessibility for clinical and home- based applications. This study highlights the transformative potential of voice- based diagnostic systems driven by machine learning. By offering a cost-effective and scalable alternative to traditional methods, the proposed system addresses critical challenges in early PD detection. Its real-time capabilities pave the way for widespread adoption in clinical and telehealth settings, ultimately enhancing early diagnosis and improving outcomes for individuals with Parkinson’s Disease.

Klasifikasi Pengucapan Huruf Hijaiyah Berbasis Android Menggunakan CNN dengan Fitur Mel-Spectrogram

Mastery of Hijaiyah letters is a fundamental basis in learning the Qur'an, but data from the IIQ Community Service Institute 2021/2022 shows that 72.25% of the 3,111 Muslims tested have not been able to read the Qur'an properly. This research aims to develop an Android-based Hijaiyah letter pronunciation classification system using Convolutional Neural Network (CNN) with mel-spectrogram features. The research methodology includes collecting 8,904 voice samples from 53 participants at Pondok Tahfidz Yanbu'ul Qur'an Menawan, preprocessing data using MFCC techniques, developing CNN models, and implementing the system in the form of mobile applications with MVVM architecture. The test results showed promising performance with some classes achieving 100% accuracy and an average overall accuracy of 83.80%, although there were challenges in some classes such as “alif_dommah” and “ghaiin_dommah” which had an accuracy below 40%. The developed system successfully provides an interactive learning platform through the integration of mobile applications with the Flask API, but still requires further development, especially in expanding the dataset to overcome overfitting problems and improve the generalization ability of the model.

Machine Learning Models With Hyperparameter Optimization for Voice Pathology Classification on Saarbrücken Voice Database.

Early diagnosis and referral are crucial in the treatment of voice disorders. Contemporary investigations have indicated the efficacy of voice pathology detection systems in significantly contributing to the evaluation of voice disorders, facilitating early diagnosis of such pathologies. These systems leverage machine learning methodologies, widely applied across diverse domains, and exhibit particular potential in the realm of voice pathology classification. However, machine learning models and performance metrics employed in these studies vary significantly, making it challenging to determine the optimal model for voice pathology classification. In this study, healthy and pathological voices were classified with state-of-the-art machine learning models, and the performance results of the models were compared. The voice samples employed in our research were sourced from the Saarbrücken Voice Database, a reputable German database. Feature extraction from voice signals was conducted using the Mel Frequency Cepstral Coefficients method. To assess and enhance the models' performance adequately, we employed hyperparameter optimization and implemented a 10-fold cross-validation approach. The outcomes revealed that the support vector machine model exhibited the highest accuracy, achieving 99.19% and 99.50% accuracies in the classification of male and female voice pathologies, respectively.

Harnessing machine learning in diagnosing complex hoarseness cases

PurposeTraditional vocal fold pathology recognition typically requires expertise of laryngologists and advanced instruments, primarily through direct visualization. This study aims to augment this conventional paradigm by introducing a parallel diagnostic procedure. Our objective is to harness a machine-learning algorithm designed to discern intricate patterns within patients' voice recordings to distinguish not only between healthy and hoarse voices but also among various specific disorders. Materials and methodsWe employed a machine-learning algorithm, utilizing transfer learning on the HuBERT model with Saarbruecken Voice Database samples. The study was conducted in two stages: a binary classifier distinguishes healthy and hoarse voices, while a subsequent multi-class classifier identifies specific voice disorders. Data from 2103 sessions, including over 25,000 components, representing diverse pathologies and healthy individuals, was analyzed. The models were trained, validated, and tested with a focus on robustness and accuracy in diagnosis. ResultsThe binary classifier achieved 82 % accuracy in distinguishing healthy from pathological voices. The multi-class algorithm which aims to identify specific laryngeal disorders obtained the highest accuracy (>93 %) for Laryngeal Dystonia. Noteworthy is the persistent challenge posed by Laryngeal Dystonia, a condition lacking a definitive diagnostic modality. ConclusionsOur findings demonstrate the feasibility of utilizing machine-learning algorithms to process voice samples, categorizing them into distinct pathologies. This approach holds potential for enhance patient triage, streamline diagnostics, and elevate overall patient care. Particularly valuable for challenging diagnoses, such as Laryngeal Dystonia, this method underscores the transformative role of machine learning in optimizing healthcare practices.

Typical forms of counteraction to the investigation of criminal offences related to interference with the activities of public authorities

The article substantiates the position that counteraction to an investigation is a complex phenomenon which may encompass various forms of behaviour implemented by interested parties with the aim of preventing detection of a criminal offence, and making it impossible to fulfil the tasks of criminal proceedings. It is found that during the investigation of criminal offences related to interference with the activities of representatives of public authorities, suspects most often implement the following forms of behaviour as ways to counteract the investigation: 1) refuse to testify; 2) give false testimony; 3) hide from pre-trial investigation authorities; 4) fail to appear when summoned by the investigator, prosecutor, investigating judge; 5) exert influence on the victim, witnesses, refuse to provide biological samples, handwriting, voice samples, etc. It is established that other interested persons at the stage of pre-trial investigation of criminal offences of the type under study most often implement the following forms of behaviour to counteract the investigation: exerting influence on participants to criminal proceedings; implementation of actions to conceal crimes, including: staging, falsification, disguise, destruction, etc.; giving false testimony; disclosure of the secret of pre-trial investigation; refusal to provide biological samples or samples of handwriting, voice, etc.; delaying the pre-trial proceedings, delaying the pre-trial investigation, for example, by ignoring the investigator's calls, prolonged familiarisation with the criminal proceedings; failure to appear when summoned; failure to fulfil procedural obligations acquired in connection with obtaining the relevant procedural status in criminal proceedings; hiding from the pre-trial investigation authorities, etc. It is also emphasised that the prospects for further research should be to identify effective ways to overcome the following forms of resistance to investigation.

Is Partial Deafness Dysphonia Gender Related?

This study aimed to examine whether voice disorders accompanying postlingual PD in adults present gender differences. The study group involved 30 patients with post lingual PD (15 females and 15 males) in whom pre existing voice disorders were excluded. Voice samples were recorded and analyzed using Multi-dimensional Voice Program by Kay Pentax. The study showed that adults with PD develop voice dysfunctions; however, no major differences between females and males were found. This study aimed to examine whether voice disorders accompanying postlingual PD in adults present gender differences. The word gender in the paper is understood as a set of biological attributes associated with physical and physiological features; therefore, is used as the equivalent of sex.

Pre-trained Convolutional Neural Networks Identify Parkinson's Disease from Spectrogram Images of Voice Samples.

Machine learning approaches including deep learning models have shown promising performance in the automatic detection of Parkinson's disease. These approaches rely on different types of data with voice recordings being the most used due to the convenient and non-invasive nature of data acquisition. Our group has successfully developed a novel approach that uses convolutional neural network with transfer learning to analyze spectrogram images of the sustained vowel /a/ to identify people with Parkinson's disease. We tested this approach by collecting a dataset of voice recordings via telephone lines, which have limited bandwidth. This study builds upon our prior results in two major ways: First, we tested the performance of our approach on a larger voice dataset recorded using smartphones with wide bandwidth. Our results show comparable performance between two datasets generated using different recording platforms where we report differences in most important features resulting from the limited bandwidth of telephonic lines. Second, we compared the classification performance achieved using linear-scale and mel-scale spectrogram images and showed a small but statistically significant gain using mel-scale spectrograms. The convolutional neural network with transfer learning approach showed superior performance against conventional machine learning methods that collapse measurements across time to generate feature vectors.

Balss pamatfrekvence afektīvā un neitrālā prosodijā dažādas strukturālās sarežģītības lingvistiskajās vienībās

Emotional states can be characterised by specific acoustic parameters. Fundamental frequency (F0) reflects biomechanical characteristics of vocal folds and is sensitive to affective components. Vocally expressed negative emotions are accompanied by increased muscular activation, while positive and neutral emotions result in fewer physiological changes. The present study aimed to research (1) whether the length of linguistic unit impacts the change of F0; (2) whether F0 differs in words, phrases, and continuous speech produced in neutral, happy, and angry prosody. Methods: Ten professional actors (5 males, 5 females) recorded nine voice samples (4 words, 4 phrases, 1 paragraph) in neutral, happy, and angry intonations. Each linguistic unit was recorded three times, resulting in a data pool of 810 samples. Six experts assessed the affective component of each actor’s performance to obtain 270 highest-ranked samples. Selected recordings were analysed using the software PRAAT v. 6.1.31. and script Vocal fundamental frequency, v. 02.04 (Phonanium, 2019). Results: In females, there were no statistically significant differences in speaking F0 between different linguistic units within a single prosody type. In males, mean F0 significantly increased with the increasing complexity of the linguistic unit in neutral, but not affective conditions. Overall, the highest speaking F0 was observed in happy prosody for females and in angry prosody for males. Both genders showed significant differences in mean F0 between neutral and affective prosody in words and phrases. A statistically significant difference in F0 between angry and happy intonation was observed in phrases in males and in text in females. Conclusion: The results showed that affective prosody did not impact mean F0 between linguistic units of different complexity. However, emotional intonation increased speaking F0. The differences in F0 between neutral and affective prosody were observed in all components of linguistic hierarchy, from single words to continuous speech. Since voice is used in daily communication, recognizing the role of emotions in vocal output can improve understanding of functional voice disorders.

Colombian College Professors Work-Related Health: Associations Between Stress and Voice Acoustics Parameters

Introduction. Teachers have a high risk of developing voice disorders and high-stress levels due to their working conditions. Moreover, stress causes changes at a physiological level in different systems such as the cardiac, gastrointestinal, and respiratory systems. In the latter, the rate of airflow is increased producing significant changes in the acoustic parameters of the voice. Methods. An exploratory, correlational, longitudinal study was conducted to investigate the association between perceived stress and three acoustic parameters related to voice perturbation and harmonicity (jitter, shimmer, and harmonics-to-noise ratio) among college professors. The study also aimed to explore potential changes in this association over the follow-up period. Twenty-four college professors participated in the study. Participants completed a questionnaire that gathered information on socio-demographic characteristics, working conditions, and stress perception. Voice samples were collected from each participant and subjected to acoustic analysis using Praat software. To examine the associations between stress levels and the acoustic parameters, generalized linear models (GLM) with a gamma distribution were employed. Results. We found that professors with low stress levels had increased jitter and shimmer values; whereas participants with moderate and high stress levels had increased harmonics-to-noise ratio values compared to those with a lower stress level. Conclusions. Stress has an important effect on voice perturbation and harmonicity parameters. These results justify the design of interdisciplinary workplace interventions for voice disorders among teachers that include activities on stress management.

Objective Voice Analysis in Partial Deafness: Comparison of Multi-Dimensional Voice Program (MDVP) and VOXplot Results.

Acoustic analysis of voice enables objective assessment of voice to diagnose changes in voice characteristics, and track the progress of therapy. In contrast to subjective assessment, objective measurements provide mathematical results referring to specific parameters and can be analyzed statistically. Changes in the voice of patients with partial deafness (PD) were not widely described in the literature, and recent studies referred to the voice parameters measured in this group of patients only using the multi-dimensional voice program (MDVP) by Kay Pentax. This paper describes the results of acoustic analysis of voice in patients with PD using VOXplot, and compares the results with those achieved with MDVP. Background/Objectives: The purpose of this study was a VOXplot objective analysis of voice in individuals with PD and to assess consistency with results obtained using MDVP and with perceptual assessment. Methods: Voice samples from 22 post-lingual PD individuals were recorded. They included continuous speech (cs) and sustained vowels (sv). The control group consisted of 22 healthy individuals with no history of voice or hearing dysfunction. The samples were analyzed with MDVP followed by VOXplot version 2.0.0 Beta. Statistical analysis was performed using a t-test paired with two samples for means. All individuals were also subjected to a perceptual voice assessment using the GRBAS by Hirano. Results: Differences were observed in 13 VOXplot parameters measured in voice samples of adults with PD compared with those in the control group. Both multiparametric indices, AVQI and ABI, showed a statistical increase. When it comes to MDVP parameters correlating with breathiness, all of them (shim dB, APQ, NHR, SPI, and NSH) increased in patients with partial deafness, reflecting a breathy voice. Only one increase in the SPI was not statistically significant. Seven MDVP parameters correlating with hoarseness were elevated, and five (Jitt%, vF0, Shim dB, APQ, and NHR) showed a statistically significant increase. Correlations were found of VOXplot and MDVP parameters with perceptual voice assessment. Conclusions: Both programs for objective assessment showed voice abnormalities in patients with PD compared with the control groups. There was a poor to moderate level of consistency in the results achieved using both systems. Correlations were also found with GRBAS assessment results.

Sensitivity of Acoustic Voice Quality Measures in Simulated Reverberation Conditions.

Room reverberation can affect oral/aural communication and is especially critical in computer analysis of voice. High levels of reverberation can distort voice recordings, impacting the accuracy of quantifying voice production quality and vocal health evaluations. This study quantifies the impact of additive simulated reverberation on otherwise clean voice recordings as reflected in voice metrics commonly used for voice quality evaluation. From a larger database of voice recordings collected in a low-noise, low-reverberation environment, voice samples of a sustained [a:] vowel produced at two different speaker intents (comfortable and clear) by five healthy voice college-age female native English speakers were used. Using the reverb effect in Audacity, eight reverberation situations indicating a range of reverberation times (T20 between 0.004 and 1.82 s) were simulated and convolved with the original recordings. All voice samples, both original and reverberation-affected, were analyzed using freely available PRAAT software (version 6.0.13) to calculate five common voice parameters: jitter, shimmer, harmonic-to-noise ratio (HNR), alpha ratio, and smoothed cepstral peak prominence (CPPs). Statistical analyses assessed the sensitivity and variations in voice metrics to a range of simulated room reverberation conditions. Results showed that jitter, HNR, and alpha ratio were stable at simulated reverberation times below T20 of 1 s, with HNR and jitter more stable in the clear vocal style. Shimmer was highly sensitive even at T20 of 0.53 s, which would reflect a common room, while CPPs remained stable across all simulated reverberation conditions. Understanding the sensitivity and stability of these voice metrics to a range of room acoustics effects allows for targeted use of certain metrics even in less controlled environments, enabling selective application of stable measures like CPPs and cautious interpretation of shimmer, ensuring more reliable and accurate voice assessments.

The doctor will polygraph you now

Artificial intelligence (AI) methods have been proposed for the prediction of social behaviors that could be reasonably understood from patient-reported information. This raises novel ethical concerns about respect, privacy, and control over patient data. Ethical concerns surrounding clinical AI systems for social behavior verification can be divided into two main categories: (1) the potential for inaccuracies/biases within such systems, and (2) the impact on trust in patient-provider relationships with the introduction of automated AI systems for “fact-checking”, particularly in cases where the data/models may contradict the patient. Additionally, this report simulated the misuse of a verification system using patient voice samples and identified a potential LLM bias against patient-reported information in favor of multi-dimensional data and the outputs of other AI methods (i.e., “AI self-trust”). Finally, recommendations were presented for mitigating the risk that AI verification methods will cause harm to patients or undermine the purpose of the healthcare system.

A comprehensive voice dataset for Hindko digit recognition.

Hindko is a language primarily spoken in Northwestern areas of Pakistan. Approximately eight million people speak the Hindko language. According to its native speakers, it is 7th largest language of Pakistan and 2nd largest language of Khyber Pakhtunkhwa. The Hazara region is the cultural hub of Hindko language. About 80% of the population in districts like Haripur, Abbotabad and Mansehra speak Hindko. The spoken content of Hindko covers a wide range of subjects, including religion, education, poetry, politics, theater, and more. Despite all this, Hindko lacks a voice recognition system that could enhance accessibility, preserve the language, and promote digital inclusion for its speakers. This paper presents a voice recognition dataset that consists of 17,597 voice samples, and is accessible to the public for academic and research purposes. The dataset consists of 20 Hindko digits ranging from 1 to 20 and all the voice samples are taken from the students and staff and faculty of Pak-Austria Fachhochschule Institute of Applied Science and Technology.

Validation of the European French Version of the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-Vf)

ObjectiveThis study aimed to validate the French adaptation of the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-Vf) for assessing voice disorders in France. The CAPE-Vf addresses limitations of the GRBAS by providing a more sensitive, standardized approach to evaluating six vocal parameters (overall severity, roughness, breathiness, strain, pitch, and loudness) on three tasks (sustained vowels, sentence reading, and spontaneous speech). The study focused on investigating the intra- and inter-rater reliability, as well as the convergent and discriminant validity of the CAPE-Vf. MethodsThirty-four dysphonic and seven euphonic native French speakers participated in the study. Thirteen speech-language pathologists from France evaluated the voice samples using both the CAPE-Vf and GRBAS tools at a one-week interval. Intra- and inter-rater reliability were calculated using intraclass correlation coefficients (ICC), while convergent and discriminant validity were measured by correlating CAPE-Vf with GRBAS and Voice Handicap Index (VHI) scores, respectively. ResultsThe CAPE-Vf showed good intra-rater reliability for overall severity (mean ICC: 0.89), strain (ICC: 0.83), and pitch (ICC: 0.88), while roughness, breathiness, and loudness exhibited moderate reliability. Inter-rater reliability was low for most parameters, except overall severity, which demonstrated good reliability (mean ICC: 0.77). Strong correlations were observed between CAPE-Vf and GRBAS Grade (mean r: 0.84), supporting its convergent validity. Moderate correlations were found for roughness, breathiness, and strain. The CAPE-Vf’s correlation with the VHI was moderate (mean r: 0.53), reflecting its discriminant validity. ConclusionThe CAPE-Vf is a valid and reliable tool for perceptual assessment of voice disorders in French-speaking populations, with stronger psychometric properties than the GRBAS, particularly for intra-rater reliability and overall severity. While inter-rater reliability was lower, qualitative feedback suggested that improvements to the protocol, particularly for pitch and loudness ratings, could enhance its clinical applicability. The findings support the CAPE-Vf as a comprehensive tool for standardized clinical voice assessment.

Using Twang and Medialization Techniques to Gain Feminine-Sounding Speech in Trans Women

ObjectivesIn this study, we introduce an intervention based on two techniques: twang and medialization. The hypothesis is that a combination of these two techniques will enable trans women to gain feminine-sounding speech without vocal strain or harm. MethodFive trans women took part in the study. A control group of five cisgender women and five cisgender men were included. A list of 14 monosyllabic words was created, where the vowel /ɑ/ was embedded in various consonant contexts. All participants were asked to read the word list three times, each time presented in a different order. The trans women read the word list before and after intervention.Acoustic analyses of fundamental frequency and the first, second, and third formant frequencies were conducted. For the perceptual analysis, 60 voice samples were selected from the entire material.Fifteen listeners were asked whether they perceived the voice samples as feminine, masculine, or uncertain. The listeners were also asked for gender judgments based on sentences read by the trans women after intervention. ResultsThe acoustic analyses revealed an increase in fundamental frequencies and first, second, and third formants after intervention for all five trans women, approaching the values of the female controls. The perceptual judgments showed that the majority of the trans women voice samples were perceived as feminine after intervention. ConclusionsBased on the acoustic analyses and the perceptual evaluations, the conclusion seems to show that the combination of the techniques twang and medialization enable the trans women to obtain feminine attribution. Nevertheless, the study is too small for generalizations. However, a take-home message is that it is appropriate to focus primarily on resonance, in addition to speaking fundamental frequency, to gain feminine-sounding speech.