Singing Voice Research Articles

Singing Voice Handicap Index-10 Minimal Clinically Important Difference: A Prospective Determination.

The Singing Voice Handicap Index-10 (SVHI-10) is a validated patient-reported outcome measure (PROM) that assesses patients' perception of handicap related to singing voice. A normative value has been established with a score ≥20 being abnormal. However, there is no defined minimal clinically important difference (MCID). This study prospectively determines the MCID of SVHI-10 among a diverse group of singers. 103 adult singers with and without voice complaints completed SVHI-10 twice, 30 days apart. MCID for the SVHI-10 was determined using distribution-based receiver-operating characteristic (ROC) curve analysis. Twenty-two men (1 transgender), 75 women (1 transgender), and 6 nonbinary individuals participated. The most frequently reported singing genres were classical (44.7%), musical theater (17.5%), and pop (10.7%). Mean initial SVHI-10 score was 13.05 (standard deviation 7.397), and mean follow-up SVHI-10 was 13.13 (7.994). There was a significant positive correlation between initial and follow-up SVHI-10 scores (r = 0.879, p < 0.001). SVHI-10 scores were significantly higher among participants who reported voice changes in the past year (p < 0.001) or sought voice treatment (p = 0.001) compared with participants who did not. SVHI-10 scores varied significantly based on singing type. The area under the ROC curve for SVHI-10 was 0.700 (p = 0.003). The SVHI-10 MCID was determined to be 9.5. An SVHI-10 score change ≥10 should be considered clinically meaningful. This definition has been missing from the literature and will improve understanding of patients' responses to treatment, which will help advance clinical care and track research outcomes. 4 Laryngoscope, 2024.

HiddenSinger: High-quality singing voice synthesis via neural audio codec and latent diffusion models

Recently, denoising diffusion models have demonstrated remarkable performance among generative models in various domains. However, in the speech domain, there are limitations in complexity and controllability to apply diffusion models for time-varying audio synthesis. Particularly, a singing voice synthesis (SVS) task, which has begun to emerge as a practical application in the game and entertainment industries, requires high-dimensional samples with long-term acoustic features. To alleviate the challenges posed by model complexity in the SVS task, we propose HiddenSinger, a high-quality SVS system using a neural audio codec and latent diffusion models. To ensure high-fidelity audio, we introduce an audio autoencoder that can encode audio into an audio codec as a compressed representation and reconstruct the high-fidelity audio from the low-dimensional compressed latent vector. Subsequently, we use the latent diffusion models to sample a latent representation from a musical score. In addition, our proposed model is extended to an unsupervised singing voice learning framework, HiddenSinger-U, to train the model using an unlabeled singing voice dataset. Experimental results demonstrate that our model outperforms previous models regarding audio quality. Furthermore, the HiddenSinger-U can synthesize high-quality singing voices of speakers trained solely on unlabeled data.

Phoneme Recognition in Korean Singing Voices Using Self-Supervised English Speech Representations

In general, it is difficult to obtain a huge, labeled dataset for deep learning-based phoneme recognition in singing voices. Studying singing voices also offers inherent challenges, compared to speech, because of the distinct variations in pitch, duration, and intensity. This paper proposes a detouring method to overcome this insufficient dataset, and applies it to the recognition of Korean phonemes in singing voices. The method started with pre-training the HuBERT, a self-supervised speech representation model, on a large-scale English corpus. The model was then adapted to the Korean speech domain with a relatively small-scale Korean corpus, in which the Korean phonemes were interpreted as similar English ones. Finally, the speech-adapted model was again trained with a tiny-scale Korean singing voice corpus for speech–singing adaptation. In the final adaptation, melodic supervision was chosen, which utilizes pitch information to improve the performance. For evaluation, the performance on multi-level error rates based on Word Error Rate (WER) was taken. Using the HuBERT-based transfer learning for adaptation improved the phoneme-level error rate of Korean speech by as much as 31.19%. Again, on singing voices by melodic supervision, it improved the rate by 0.55%. The significant improvement in speech recognition underscores the considerable potential of a model equipped with general human voice representations captured from the English corpus that can improve phoneme recognition on less target speech data. Moreover, the musical variation in singing voices is beneficial for phoneme recognition in singing voices. The proposed method could be applied to the phoneme recognition of other languages that have less speech and singing voice corpora.

Perceived Benefits of Interdisciplinary Coaching Strategies for Professional Vocalists: A Pilot Study

ABSTRACT Professional vocalists, specifically teachers, are at increased risk of missing work and having vocal damage due to vocal misuse and overuse. The purpose of this pilot study was to examine the perceptions of participants (N = 4) taking part in interdisciplinary voice coaching sessions with both a singing and theatre voice specialist. Participant perceptions were obtained using the Voice Handicap Index (VHI), individual weekly journals, and a post-study researcher-created questionnaire. Results are discussed in terms of potential for future research and application of results to mutli-disciplinary voice training options for professional voice users as a solution to mitigate voice problems.

Three-Dimensional Finite Element Modeling of the Singer’s Formant Cluster Optimization by Epilaryngeal Narrowing With and Without Velopharyngeal Opening

This study aimed to find the optimal geometrical configuration of the vocal tract to increase the total acoustic energy output of human voice in the frequency interval 2-3.5 kHz “singer’s formant cluster”, (SFC) for vowels [a:] and [i:] considering epilaryngeal changes and the velopharyngeal opening (VPO). The study applied 3D volume models of the vocal and nasal tract based on computer tomography (CT) images of a female speaker. The epilaryngeal narrowing (EN) increased the total sound pressure level (SPL) and SPL of the SFC by diminishing the frequency difference between acoustic resonances F3 and F4 for [a:] and between F2 and F3 for [i:]. The effect reached its maximum at the low pharynx/epilarynx cross-sectional area ratio 11.4:1 for [a:] and 25:1 for [i:]. The acoustic results obtained with the model optimization are in good agreement with the results of an internationally recognized operatic alto singer. With the EN and the VPO, the vocal tract input reactance was positive over the entire fo singing range (ca. 75-1500 Hz). The VPO increased the strength of the SFC and diminished the SPL of F1 for both vowels, but with EN the SPL decrease was compensated. The effect of EN is not linear and depends on the vowel. Both the EN and the VPO alone and together can support (singing) voice production.

Singing voice separation based on U-NET neural network and its application

Singing Voice separation is a very important method that helps people to obtain a vocal or accompaniment for further usage like music composing. In this paper, I developed a Song accompaniment separation application based on Convolutional neural network Unet model and Deep learning (DL). Basically, the U-net convolutional network is applied to the two-dimensional spectrum of a music audio in order to achieve a more precise separation effect. During this deep learning process, a significant amount of data was inserted into the model to get the best fit parameters. Then, after all, the model can automatically generate a mask that filters out the accompaniment or vocal part that users want.. This completed model is then transmitted to the terminal of a internet server to enable users to get access to the model through internet. After the test of both quantitative evaluation and subjective assessment, the application impressively achieved the Song accompaniment Separation that provides a good quality of both vocal and song.

RealSinger: Ultra-realistic singing voice generation via stochastic differential equations

Synthesizing high-quality singing voice from music score is a challenging problem in music generation and has many practical applications. Samples generated by existing singing voice synthesis (SVS) systems can roughly reflect the lyrics, pitch and duration in a given score, but they fail to contain necessary details. In this paper, based on stochastic differential equations (SDE) we propose RealSinger to generate 22.05kHz ultra-realistic singing voice conditioned on a music score. Our RealSinger learns to find the stochastic process path from a source of white noise to the target singing voice manifold under the conditional music score, allowing to sing the music score while maintaining the local voice details of the target singer. During training, our model learns to accurately predict the direction of movement in the ambient Euclidean space onto the low-dimensional singing voice manifold. RealSinger’s framework is very flexible. It can either generate intermediate feature representations of the singing voice, such as mel-spectrogram, or directly generate the final waveform, as in the end-to-end style which rectify defects and accumulation errors introduced by two-stage connected singing synthesis systems. An extensive subjective and objective test on benchmark datasets shows significant gains in perceptual quality using RealSinger. The mean opinion scores (MOS) obtained with RealSinger are closer to those of the human singer’s original high-fidelity singing voice than to those obtained with any state-of-the-art method. Audio samples are available at https://realsinger.github.io/.

Automatic Lyric Transcription and Automatic Music Transcription from Multimodal Singing

Automatic lyric transcription (ALT) refers to transcribing singing voices into lyrics, while automatic music transcription (AMT) refers to transcribing singing voices into note events, i.e., musical MIDI notes. Despite these two tasks having significant potential for practical application, they are still nascent. This is because the transcription of lyrics and note events solely from singing audio is notoriously difficult due to the presence of noise contamination, e.g., musical accompaniment, resulting in a degradation of both the intelligibility of sung lyrics and the recognizability of sung notes. To address this challenge, we propose a general framework for implementing multimodal ALT and AMT systems. Additionally, we curate the first multimodal singing dataset, comprising N20EMv1 and N20EMv2, which encompasses audio recordings and videos of lip movements, together with ground truth for lyrics and note events. For model construction, we propose adapting self-supervised learning models from the speech domain as acoustic encoders and visual encoders to alleviate the scarcity of labeled data. We also introduce a residual cross-attention mechanism to effectively integrate features from the audio and video modalities. Through extensive experiments, we demonstrate that our single-modal systems exhibit state-of-the-art performance on both ALT and AMT tasks. Subsequently, through single-modal experiments, we also explore the individual contributions of each modality to the multimodal system. Finally, we combine these and demonstrate the effectiveness of our proposed multimodal systems, particularly in terms of their noise robustness.

Population characteristics and self-assessment of speaking and singing voice in Polish contemporary commercial music singers-an exploratory, cross-sectional study.

The domination of the Contemporary Commercial Music (CCM) industry in music markets has led to a significant increase in the number of CCM performers. Performing in a wide variety of singing styles involves exposing CCM singers to specific risk factors potentially leading to voice problems. This, in turn, necessitates the consideration of this particular group of voice users in the Occupational Health framework. The aim of the present research was threefold. First, it sought to profile the group of Polish CCM singers. Second, it was designed to explore the prevalence of self-reported voice problems and voice quality in this population, in both speech and singing. Third, it aimed to explore the relationships between voice problems and lifetime singing involvement, occupational voice use, smoking, alcohol consumption, vocal training, and microphone use, as potential voice risk factors. The study was conducted in Poland from January 2020 to April 2023. An online survey included socio-demographic information, singing involvement characteristics, and singers' voice self-assessment. The prevalence of voice problems was assessed by the Polish versions of the Vocal Tract Discomfort Scale (VTDS) and the Singing Voice Handicap Index (SVHI). Also, a self-reported dysphonia symptoms protocol was applied. The perceived overall voice quality was assessed by a Visual Analogue Scale (VAS) of 100 mm. 412 singers, 310 women and 102 men, completed the survey. Nearly half of the studied population declared lifetime singing experience over 10 years with an average daily singing time of 1 or 2 h. 283 participants received vocal training. For 11.4% of respondents, singing was the primary income source, and 42% defined their career goals as voice-related. The median scores of the VTDS were 11.00 (0-44) and 12.00 (0-40) for the Frequency and Severity subscales, respectively. The median SVHI score of 33 (0-139) was significantly higher than the normative values determined in a systematic review and meta-analysis (2018). Strong positive correlations were observed between SVHI and both VTD subscales: Frequency (r = 0.632, p < 0.001) and Severity (r = 0.611, p < 0.001). The relationships between most of the other variables studied were weak or negligible. The examined CCM singers exhibited substantial diversity with regard to musical genre preferences, aspirations pertaining to singing endeavors, career affiliations, and source of income. Singing voice assessment revealed a greater degree of voice problems in the examined cohort than so far reported in the literature, based on the SVH and VTDS.

Fat Liberation in the Singing Voice Studio

ABSTRACT In recent years, creating a sense of belonging for marginalized identities within singing voice pedagogy has increased in both visibility and cultural importance. Support and resources on the practice of cultural humility have bolstered the role of social justice in singing voice studios and called attention to the link between affirming students’ identities and their ability to access their fullest artistic expression. However, body size is still often left out of conversations about social justice. If teachers wish to be considered progressive and inclusive, there is a duty to lead the charge for fat liberation within singing voice pedagogy. Through the voices of 66 singers who identify as fat, this article details how anti-fat bias can manifest in the voice studio and identifies common pedagogical practices that have historically failed to serve fat singers. Underpinned by data from 66 interviews, recommendations for dismantling implicit anti-fat bias, working inclusively within the studio and beyond, and advocating for systemic change will be made. The article concludes with specific guidance on how to become a fat liberationist so that students of every body size may find a true sense of belonging in the singing voice studio.

Perceptions Around Vocal Health Among Group Fitness Instructors and Pilot Use of Group Fitness Instructor–Specific Voice Handicap Questionnaire

Purpose: This study aimed to examine the vocal behaviors of group fitness instructors (GFIs), explore awareness among GFIs about voice-specialized health care, and assess the potential utility of a GFI-specific voice handicap questionnaire. Method: An anonymous 27-question survey was distributed digitally to GFIs. In addition to collecting demographic data and information about vocal behaviors and demands, the survey included a 10-question GFI-specific questionnaire (GFI-VHI-10) adapted from the Singing Voice Handicap Index-10. Results: Fifty-two GFIs participated in the study (75% female, 25% male) with a mean age of 44.6 years. Forty-six percent reported having a voice problem (VP) since entering the fitness industry. Of those 24 individuals, only four received treatment. There was a statistically significant difference in mean GFI-VHI-10 scores between the VP group (11.87) and the non-VP group (6.1), ( p &lt; .001). Of all respondents, 56% reported never or almost never wearing a microphone while instructing. When asked what they would do if they experienced a VP that lasted for longer than 2 weeks, only 9.6% would see a laryngologist, 23% would see an otolaryngologist, 31% would see their primary care provider, 11.5% were unsure, and 25% would ignore it and hope for spontaneous improvement. Conclusions: Although GFIs are at increased risk for vocal health challenges due to their occupational voice demands, their adherence to vocal health maintenance behaviors and awareness of voice-specialized health care may be limited. The GFI-VHI-10 is a promising tool in that it captured higher scores in the VP group than the non-VP group; however, further exploration is warranted.

Effectiveness of a Voice Training Program for Religious Officials in Turkey.

This study examines the effectiveness of a voice training program designed for Islamic religious officials, who are occupational voice users with a significant vocal load. The participants included 34 healthy religious officials whose acoustic measures were within normal voice ranges for healthy adults (jitter < 1%; shimmer < 3%) and reported no voice complaints. Participants were randomly divided into two groups (experimental, n = 17; control, n = 17). The two-stage voice training program consisted of 32 sessions over 8 weeks with informative and voice exercise stages. Objective and subjective voice measurements were performed at the beginning and end of the research. Objective measurements included fundamental frequency, percentage of vocal pitch perturbation (jitter), percentage of vocal intensity perturbation (shimmer), and harmonics-to-noise ratio. Subjective voice measurements included the Singing Voice Handicap Index, Vocal Fatigue Index (VFI), and Voice-Related Quality of Life (V-RQoL) scores. All initial measurements other than VFI scores were within acceptable limits for both groups. There were no significant differences between the groups initially (p > 0.05) and no significant changes in the control group in the second evaluation (p > 0.05). However, there was significant improvement in the experimental group after the training program in all measures, including VFI scores (p < 0.05). This study shows the positive results of a voice training program. Voice training should be integrated into the formal education of occupational voice users or in-service training programs of relevant institutions.

Quebec French Translation, Cultural Adaptation, and Validation of the Singing Voice Handicap Index-10 Questionnaire for Singers with Dysphonia

OBJECTIVES/HYPOTHESISDysphonia is a common voice disorder that can significantly impact a person’s life; it requires a collaborative evaluation by both speech-language pathologists and otolaryngologists that takes the patient’s perspective into account. The aim of this study was to translate and culturally adapt the Singing Voice Handicap Index questionnaire (SVHI-10), a reliable patient-reported outcome evaluation tool for dysphonia, for the Quebec French population. The result is the Singing Voice Handicap Index-10-QC (SVHI-10-QC). STUDY DESIGNThis study was a prospective translation and validation process. METHODSThe translation process complied with international recommendations and followed a standard forward-backward translation procedure and cognitive debriefing with ten singers. The Quebec French version was administered to two study samples: 30 vocal professionals with no dysphonia and 53 vocal professionals with dysphonia as one of their primary complaints. The SVHI-10-QC was assessed for construct validity, internal consistency, discriminatory capacity, and test-retest reliability. RESULTSThe SVHI-10-QC is valid, reliable, and ready for use with singer-patients suffering from dysphonia. CONCLUSIONSThe SVHI-10-QC is a reliable and valid tool for assessing the impact of dysphonia on French Quebec singers’ quality of life.

StyleSinger: Style Transfer for Out-of-Domain Singing Voice Synthesis

Style transfer for out-of-domain (OOD) singing voice synthesis (SVS) focuses on generating high-quality singing voices with unseen styles (such as timbre, emotion, pronunciation, and articulation skills) derived from reference singing voice samples. However, the endeavor to model the intricate nuances of singing voice styles is an arduous task, as singing voices possess a remarkable degree of expressiveness. Moreover, existing SVS methods encounter a decline in the quality of synthesized singing voices in OOD scenarios, as they rest upon the assumption that the target vocal attributes are discernible during the training phase. To overcome these challenges, we propose StyleSinger, the first singing voice synthesis model for zero-shot style transfer of out-of-domain reference singing voice samples. StyleSinger incorporates two critical approaches for enhanced effectiveness: 1) the Residual Style Adaptor (RSA) which employs a residual quantization module to capture diverse style characteristics in singing voices, and 2) the Uncertainty Modeling Layer Normalization (UMLN) to perturb the style attributes within the content representation during the training phase and thus improve the model generalization. Our extensive evaluations in zero-shot style transfer undeniably establish that StyleSinger outperforms baseline models in both audio quality and similarity to the reference singing voice samples. Access to singing voice samples can be found at https://stylesinger.github.io/.

“The Full Monty”: Taking a Fully Qualitative Approach to Research By Singers, With Singers, For Singers

ABSTRACT This article extends the scholarly conversation on research within performing arts voice disciplines by exploring how “fully qualitative” research can help bridge the gap between rigorous and relevant research. Fully qualitative research is informed by a qualitative sensibility which values the subjectivity of researcher and participant, and encourages reflexivity, transparency, and criticality throughout the research process. Taking singing voice research as an example, the article presents a reflective review of trends in this field, identifying the dominance of the scientific approach. Fully qualitative research is presented as a way to gain new insights into singing which are unlikely to flow from research using the scientific method. The article outlines how a fully qualitative approach by singers to conducting rigorous research with singers can produce findings which are relevant for singers. While illustrative examples are presented from singing voice research, voice practitioners and researchers from other disciplines are encouraged to consider a fully qualitative approach in their research and to develop an awareness of quality indicators. Fully qualitative voice research complements existing scientific knowledge and has the potential to make significant contributions to our understanding of the role of voice in human life.

3 directional Inception-ResUNet: Deep spatial feature learning for multichannel singing voice separation with distortion.

Singing voice separation on robots faces the problem of interpreting ambiguous auditory signals. The acoustic signal, which the humanoid robot perceives through its onboard microphones, is a mixture of singing voice, music, and noise, with distortion, attenuation, and reverberation. In this paper, we used the 3D Inception-ResUNet structure in the U-shaped encoding and decoding network to improve the utilization of the spatial and spectral information of the spectrogram. Multiobjectives were used to train the model: magnitude consistency loss, phase consistency loss, and magnitude correlation consistency loss. We recorded the singing voice and accompaniment derived from the MIR-1K dataset with NAO robots and synthesized the 10-channel dataset for training the model. The experimental results show that the proposed model trained by multiple objectives reaches an average NSDR of 11.55 dB on the test dataset, which outperforms the comparison model.

Retracted: The New Approach Research on Singing Voice Detection Algorithm Based on Enhanced Reconstruction Residual Network

Retracted: The New Approach Research on Singing Voice Detection Algorithm Based on Enhanced Reconstruction Residual Network

A hybrid CNN-LSTM model with adaptive instance normalization for one shot singing voice conversion

&lt;abstract&gt;&lt;p&gt;Singing voice conversion methods encounter challenges in achieving a delicate balance between synthesis quality and singer similarity. Traditional voice conversion techniques primarily emphasize singer similarity, often leading to robotic-sounding singing voices. Deep learning-based singing voice conversion techniques, however, focus on disentangling singer-dependent and singer-independent features. While this approach can enhance the quality of synthesized singing voices, many voice conversion systems still grapple with the issue of singer-dependent feature leakage into content embeddings. In the proposed singing voice conversion technique, an encoder decoder framework was implemented using a hybrid model of convolutional neural network (CNN) accompanied by long short term memory (LSTM). This paper investigated the use of activation guidance and adaptive instance normalization techniques for one shot singing voice conversion. The instance normalization (IN) layers within the auto-encoder effectively separated singer and content representations. During conversion, singer representations were transferred using adaptive instance normalization (AdaIN) layers. This singing voice system with the help of activation function prevented the transfer of singer information while conveying the singing content. Additionally, the fusion of LSTM with CNN can enhance voice conversion models by capturing both local and contextual features. The one-shot capability simplified the architecture, utilizing a single encoder and decoder. Impressively, the proposed hybrid CNN-LSTM model achieved remarkable performance without compromising either quality or similarity. The objective and subjective evaluation assessments showed that the proposed hybrid CNN-LSTM model outperformed the baseline architectures. Evaluation results showed a mean opinion score (MOS) of 2.93 for naturalness and 3.35 for melodic similarity. These hybrid CNN-LSTM techniques allowed it to perform high-quality voice conversion with minimal training data, making it a promising solution for various applications.&lt;/p&gt;&lt;/abstract&gt;

Associations of postural activities and knowledge for voice with breathing issues and voice-physical-disorders among lyric singers

Objectives/HypothesisThe purpose of this research was to study the associations of postural activities and knowledge of the voice of opera singers, recognized in the literature for the specificity of their posture. Additionally, the link between vocal pathologies and body pains on one hand, and posture control on the other hand was investigated. MethodsA questionnaire including 90 questions was distributed to singers in France and overseas during 6 months. Ninety-eight opera singers participated in the survey. Data were analyzed using Excel and Stata software. ResultsThe results showed that the singers who paid more attention to posture, postural work while singing, had knowledge of anatomy and postural role in singing voice, healthy lifestyle and good body habits, had less vocal discomfort and pathologies, a better vocal preservation and used more costo-abdominal breathing. On the other hand, knowledge of postural role and postural work was linked with pain and vocal fatigue. Furthermore, a preventive need was revealed concerning the providing of knowledge and skills to singers, but also to singing teachers, speech therapists and stage workers. ConclusionOur study shows that benefits were found in the voice quality of opera singers with anatomical and postural knowledge and who work on posture as part of their vocal work compared to others. Postural work and knowledge increase attention to body pain and vocal fatigue. These results can inform health care providers, opera singers and their teachers and performers of the benefits of posture on operatic voice quality. Accordingly, this study sparks new ideas for postural work and therapy in lyric voice.

Autopercepção do Impacto Imediato na Voz de Cantores Gospel após uma Apresentação de uma hora

ABSTRACT Purpose To investigate the immediate impact on the voice of gospel singers with and without vocal complaints after a one-hour individual presentation. Methods Application of an online questionnaire that addressed the following aspects: 1 - Sociodemographic data; 2 – Self-assessment of the ability to sing using the Evaluation of the Ability to Sing Easily (EASE-BR) protocol; 3 - Self-assessment of vocal fatigue symptoms using the Vocal Fatigue Index (VFI) protocol; and 4 - Self-assessment of voice handicap using the Voice Handicap Index 10 (VHI-10) protocol. Participants were divided into two groups: Group with Vocal Complaint (WVC) and Group with no Vocal Complaint (WnVC) based on the total score of the IDV-10. Data underwent descriptive and inferential statistical analysis with a significance level of 5%. Results The study included 43 gospel singers with a median age of 34 years: 32 were in the WnVC group and 11 were in the WVC group. The WVC group reported hoarseness and experienced more difficulty while singing in the EASE, resulting in higher scores in both the VHI-10 and VFI protocols. A positive correlation was observed between singing difficulty and vocal handicap due to fatigue in amateur gospel singers, with this correlation being stronger within the WnVC group. Conclusion After one hour of performance, singers with vocal complaints exhibited higher rates of vocal fatigue, vocal disadvantage, and greater difficulty in singing. Singers without complaints may have their ability to sing impaired by vocal fatigue. Variations in singing ability and vocal handicaps in amateur gospel singers may be directly related to vocal fatigue.