Voice Samples Research Articles

Treatment Outcomes of Type 1 Thyroplasty Using Gore-Tex® Following Injection Laryngoplasty.

To compare voice-related outcomes of type 1 thyroplasty using Gore-Tex® between patients with and without augmentation injection laryngoplasty (IL) prior to surgery. Forty-five patients who underwent Gore-Tex® thyroplasty at a single institution by a single surgeon between November 2016 and February 2019 were identified as those who previously had IL (n = 20) and those without IL (n = 25). Pre- and post-operative voice-related primary outcomes were evaluated using the GRBAS, and CAPE-V auditory-perceptual rating scales and secondary outcome were evaluated using the VRQOL. Pre- and post-operative voice samples were blinded, randomized, and analyzed by 3 voice-specialized speech pathologists to obtain CAPE-V scores. The VRQOL and GRBAS scores were obtained from retrospective chart review. Student's t test with a paired one-tailed distribution was used for comparisons within groups and 2-sample equal variance for comparisons between groups. Intraclass correlation coefficient determined interrater agreement. GRBAS, and VRQOL significantly improved post Gore-Tex® thyroplasty. There was no difference in improvement between patients who received pre-surgery IL and those who did not in either GRBAS or VRQOL scores, but CAPE-V showed significant improvement in the IL group. A strongly positive correlation was demonstrated between the severity of CAPE-V pre-op score and the overall improvement following surgery for both groups combined. Patients with vocal fold paralysis have a significantly better voice after Gore-Tex® thyroplasty by self-report (VRQOL) and assessment by trained voice professionals (GRBAS). Having IL prior to surgery does not adversely affect later surgical outcomes. This paper represents one of the largest analyses of voice quality outcomes of Gore-Tex® thyroplasty using validated patient scales and randomized blinded analyses.

Evaluating voice samples as a potential source of information about personality

Speech is a powerful medium through which a variety of psychologically relevant phenomena are expressed. Here we take a first step in evaluating the potential of using voice samples as non-self-report measures of personality. In particular, we examine the extent to which linguistic and vocal information extracted from semi-structured vocal samples can be used to predict conventional measures of personality. We extracted 94 linguistic features (using Linquistic Inquiry Word Count, 2015) and 272 vocal features (using pyAudioAnalysis) from 614 voice samples of at least 50 words. Using a two-stage, fully automatable machine learning pipeline we evaluated the extent to which these features predicted self-report personality scales (Big Five Inventory). For comparison purposes, we also examined the predictive performance of these voice features with respect to depression, age, and gender. Results showed that voice samples accounted for 10.67 % of the variance in personality traits on average and that the same samples could also predict depression, age, and gender. Moreover, the results reported here provide a conservative estimate of the degree to which features derived from voice samples could be used to predict personality traits and suggest a number of opportunities to optimize personality prediction and better understand how voice samples carry information about personality.

언어재활사의 주관적 음성피로도와 관련된 음향적 특성*

In addition to administering a questionnaire (J-survey), which questions individuals on subjective vocal fatigue, voice samples were collected before and after speech-language pathology sessions from 50 female speech-language pathologists in their 20s and 30s in the Daejeon and Chungnam areas. We identified significant differences in Korean Vocal Fatigue Index scores between the fatigue and non-fatigue groups, with the most prominent differences in sections one and two. Regarding acoustic phonetic characteristics, both groups showed a pattern in which low-frequency band energy was relatively low, and high-frequency band energy was increased after the treatment sessions. This trend was well reflected in the low-to-high ratio of vowels, slope LTAS, energy in the third formant, and energy in the 4,000-8,000 Hz range. A difference between the groups was observed only in the vowel energy of the low-frequency band (0-4,000 Hz) before treatment, with the non-fatigue group having a higher value than the fatigue group. This characteristic could be interpreted as a result of voice abuse and higher muscle tonus caused by long-term voice work. The perturbation parameter and shimmer local was lowered in the non-fatigue group after treatment, and the noise-to-harmonics ratio (NHR) was lowered in both groups following treatment. The decrease in NHR and the fall of shimmer local could be attributed to vocal cord hypertension, but it could be concluded that the effective voice use of speech-language pathologists also contributed to this effect, especially in the non-fatigue group. In the case of the non-fatigue group, the rhamonics-to-noise ratio increased significantly after treatment, indicating that the harmonic structure was more stable after treatment.

A Study on Voice Measures in Patients With Alzheimer's Disease

As Alzheimer's disease (AD) might provoke certain nerve disorders, patients with AD can acquire sensorimotor adaptation problems, and thus the acoustic characteristics of the speech they produce may differ from those of healthy subjects. This study aimed to (1) extract acoustic characteristics (relating to articulatory gestures) potentially useful for detecting AD and (2) examine whether these characteristics could help identify AD patients. A total of 50 individuals participated in the study, including the AD group (17 cases), the Neurologically Healthy (NH) group (13 cases), the Mild Cognitive Impairment (MCI) group (11 cases), and the Vascular Cognitive Impairment (VCI) group (9 cases). Voice samples involving three vowels (/i/, /a/, and /u/) and six consonants (/p/, /pʰ/, /t/, /tʰ/, /k/, and /kʰ/) were collected using a digital recorder (TASCAM DR40X). Microphone-to-mouth distance was maintained at 30 cm. Acoustic measures included F0, jitter, shimmer, HNR, F1, F2, F3, and VOT. One-way ANOVA tests were carried out to compare the acoustic measures among the four groups. F3 of vowel /u/, F2 bandwidth of vowel /a/, VOT of consonant /t/, and male participants' F0 of three vowels (/a/, /i/, and /u/) were found significantly different, while no significant differences were found in the other measures. Some acoustic characteristics can indeed help detect AD patients.

Voice Cloning in Real Time

Abstract: Deep learning models for natural voice cloning methods were first developed in 2016, and since then, the researchers' main attention has been on making the voice more realistic and obtaining the output voice in real time. Previously it used to take many hours of voice samples to clone a few seconds. It was decreased to a few seconds after utilizing deep learning models. We shall look at various voice cloning techniques in this paper. Multi-speaker generative models, speaker adoption, speaker encoding, vector quantization, and other techniques are among them

Objective and subjective voice evaluation in Covid 19 patients and prognostic factors affecting the voice

Coronavirus disease 2019 (COVID-19), a respiratory and systemic zoonosis form caused by a virus belonging to the Coronaviridae family. Although several studies have shown the otolaryngology symptoms are affected in COVID-19 patients, the number of studies regarding the COVID-19 effects on voice is limited. Our study aims to evaluate the effect of COVID-19 on voice objectively - subjectively and compare it with the control group. 50 hospitalized patients with laboratory-confirmed COVID-19 and 50 healthy individuals were included in the study as study and control group, respectively. All subjects were trained to vocalize a continuous/a/ vocal pattern at speech sound intensity for Maximum Phonation Time. Voice samples were recorded using a Sony (ICD-PX470) audio recorder and analyzed by the Praat program. Dysphonia grades were ranked on 4-point scales (grade: 0=none; 1= mild; 2= moderate; 3=severe). It is seen from the results that, there were significant differences between the male and female participants in acoustic parameters of fundamental frequency (F0) (p<0.001), shimmer and mean harmonic to noise ratio (HNR) (p=0.011). There was also a significant difference in F0 values of infected and healthy participants (p=0.008). However, there was no significant interaction between gender and health status in any acoustic parameters (p>0.05). The degree of thoracic computed tomography (CT) involvement had no significant effect on parameters (p>0.05), while there was a weak positive relationship between the duration of hospitalization and F0 (rs=0.397, p=0.004). Dysphonia was positively associated with health status (rs=0. 682, p<0.001), and female infected participants reported more frequent dysphonia than males. In our study, we examined the effect of COVID-19 on voice both objectively and subjectively and evaluated the relationship between CT involvement and duration of hospitalization, which made our study more reliable. Future studies with larger and more specific patient groups to investigate the relationship between COVID-19 and dysphonia will shed a light on the subject.

Analysis of the Microcontroller Resources Using Specifics for Speech Recognition

The use of neural networks for information recognition, in particular, voice, expands the functional capabilities of embedded systems on microcontrollers. But it is necessary to take into account the limitations of the microcontroller resources. The purpose of the work is to analyze the impact of voice processing parameters and neural network architecture on the degree of microcontroller resources usage. To do this, a database of samples of the keyword, samples of other words and voices, and samples of noise are created, the probability of recognizing the keyword among other words and noises is evaluated, the dependence of the amount of memory used on the microcontroller and the decision-making time on the number MFC coefficients is established, the dependence of the amount of used memory of the microcontroller and the decision-making time on the type of convolutional neural network is established also. During the experiment, the Arduino Nano 33 BLE Sense development board was used. The neural network model was built and trained on the Edge Impulse software platform. To conduct the experiment, three groups of data with the names "hello", "unknown", "noise" were created. The group "hello" contains 94 examples of the word "hello" in English, spoken by a female voice. The "unknown" group contains 167 examples of other words pronounced by both female and male voices. The "noise" group contains 166 samples of noise and random sounds. According to Edge Impulse's recommendation, 80% of the samples from each of the data groups were used to train the neural network model, and 20% of the samples were used for testing. Analysis of the results shows that with an increase in the number of MFC coefficients and, accordingly, the accuracy of keyword recognition, the amount of program memory occupied by the code increases by 480 bytes (less than 1%). For the nRF52840 microcontroller, this is not a significant increase. The amount of RAM used during the experiment did not change. Although the calculation time of the accuracy of the code word definition increased by only 14 ms (less than 5%) with the increase in the number of MFC coefficients, the calculation procedure is quite long (approximately 0.3 s) compared to the sound sample length of 1 s. This can be a certain limitation when processing a sound signal with 32-bit microcontrollers. To analyze phrases or sentences, it is necessary to use more powerful microcontrollers or microprocessors. Based on the results of experimental research, it can be stated that the computing resources of 32-bit microcontrollers are quite sufficient for recognizing voice commands with the possibility of pre-digital processing of the sound signal, in particular, the use of low-frequency cepstral coefficients. The selection of the number of coefficients does not significantly affect the amount of used FLASH and RAM memory of the nRF52840 microcontroller. The comparison results show the superiority of the 2D network in the accuracy of the keyword definition for both 12 and 13 MFC coefficients. The use of a one-dimensional convolutional neural network for voice sample recognition in the conducted experiment provides memory savings of approximately 5%. The quality of keyword recognition with the number of MFC coefficients of 12 is approximately 0.7. For 17 MFC coefficients, the recognition quality is already 0.97. The amount of RAM used in the case of the 2D network has decreased slightly. Voice sample processing time for both types of networks is practically the same. Thus, 1D convolutional neural networks have certain advantages in microcontroller applications for voice processing and recognition. The limitation of voice recognition on the microcontroller is the sufficiently long processing time of the sound sample (approximately 0.3 s) with the duration of the sample itself being 1 s, this can be explained by a sufficiently low clock frequency of 64 MHz. Increasing the clock frequency will reduce the calculation time.

Profiles and predictors of onset based differences in vocal characteristics of adults with auditory neuropathy spectrum disorder (ANSD)

PurposeOnset-based differences are understudied in Auditory Neuropathy Spectrum Disorder (ANSD) in dimensions such as voice, which is addressed in the study. The study aimed to profile and predict the best metrics of onset-related differences in acoustic vocal characteristics of early and late-onset ANSD patients. Methods31 participants (15 early and 16 late-onset) aged 15–30 years diagnosed with ANSD were included in the study. The sustained phonation of vowel /i/ recorded by the participants using android based smartphones of selected configuration was sent over email to the experimenter. Acoustic parameters (fundamental frequency, harmonic frequencies, jitter, shimmer, harmonic-to-noise ratio, cepstral peak prominence -CPP, and pitch sigma) were analysed using Praat software. ResultsResults revealed significantly increased (p < 0.05) fundamental frequency along with decreased F2 and F3 of /i/ in the early-onset ANSD compared to the late-onset group, which can be explained based on differences in the pathophysiology of the disorder. Although not statistically significant, mean perturbations (jitter and shimmer), harmonic-to-noise ratio, cepstral peak prominence, and pitch sigma were more affected in the early-onset group, reflective of lowered auditory feedback and periodicity in their voice samples. Results of discriminant analysis marked the emergence of F2, F3, and CPP as the most sensitive metrics for onset-based group differences in voice characteristics. ConclusionsThe findings from the study highlight the role of acoustical voice evaluation (especially CPP, F2 & F3) in verifying the onset of ANSD disorder. The insights from the onset-based differences seen in vocal characteristics can indirectly help audiologists in deciding the management options for ANSD.

A MFCC-CNN BASED VOICE AUTHENTICATION SECURITY

In this paper, a novel architecture is proposed using a convolutional neural network (CNN) and mel frequency cepstral coefficient (MFCC) to identify the speaker in a noisy environment. This architecture is used in a text-independent setting. The most important task in any text-independent speaker identification is the capability of the system to learn features that are useful for classification. We are using a hybrid feature extraction technique using CNN as a feature extractor combined with MFCC as a single set. For classification, we used a deep neural network which shows very promising results in classifying speakers. We made our dataset containing 60 speakers, each speaker has 4 voice samples. Our best hybrid model achieved an accuracy of 87.5%. To verify the effectiveness of this hybrid architecture, we use parameters such as accuracy and precision.

Individual Voice Dimensions' Prediction of Overall Dysphonia Severity on Two Auditory-Perceptual Scales.

Auditory-perceptual evaluation of dysphonic voice is an essential clinical activity that characterizes the nature of dysphonia and aids in planning its clinical management. Although there are multidimensional acoustic measures that correlate well with overall severity ratings, they tend to include measures that have only small or moderate correlations with individual voice characteristics frequently perceptually measured (e.g., breathiness or roughness). Given this difference between perceptual and acoustic measures, it is unclear how much individual voice characteristics contribute to a listener's perception of overall severity of dysphonia. The purpose of this study was to explore individual voice characteristics' relative contribution to the rating of overall dysphonia severity and to explore sex-related differences. Two hundred ninety-six voice samples were accessed from the Perceptual Voice Qualities Database. Roughness, breathiness, asthenia, strain, pitch, and loudness ratings from the Grade, Roughness, Breathiness, Asthenia, Strain and Consensus Auditory-Perceptual Evaluation of Voice scales were used to predict overall voice quality severity in linear regression with bootstrapped coefficients. Roughness, breathiness, and strain were the strongest predictors of overall severity. Asthenia and, to a lesser extent, pitch were also significant predictors of overall severity. Loudness was not a significant predictor. There were several sex-related differences noted, as well as differences related to the scale used. Breathiness, roughness, and strain were all important predictors of overall severity for all regressions. Clinicians should be aware of scale-related differences if they are using auditory-perceptual measures to choose voice therapy targets. Analyses accounting for perceptual strategy differences were recommended for future studies.

Deep learning applications in telerehabilitation speech therapy scenarios

Nowadays, many application scenarios benefit from automatic speech recognition (ASR) technology. Within the field of speech therapy, in some cases ASR is exploited in the treatment of dysarthria with the aim of supporting articulation output. However, in presence of atypical speech, standard ASR approaches do not provide any reliable result in terms of voice recognition due to main issues, including: (i) the extreme intra and inter-speakers variability of the speech in presence of speech impairments, such as dysarthria; (ii) the absence of dedicated corpora containing voice samples from users with a speech disability to train a state-of-the-art speech model, particularly in non-English languages. In this paper, we focus on isolated word recognition for native Italian speakers with dysarthria and we exploit an existing mobile app to collect audio data from users with speech disorders while they perform articulation exercises for speech therapy purposes. With this data availability, a convolutional neural network has been trained to spot a small number of keywords within atypical speech, according to a speaker dependent method. Finally, we discuss the benefits of the trained ASR system in tailored telerehabilitation contexts intended for patients with dysarthria who can follow treatment plans under the supervision of remote speech language pathologists.

Do the Nonlinear Dynamic Acoustic Measurements, Nonlinear Energy Difference Ratio and Spectrum Convergence Ratio, Correlate with Perceptual Evaluation of Esophageal Voice Speakers?

The acoustic assessment of phonation after total laryngectomy is challenged by signal aperiodicity which makes frequency-based acoustic measures less reliable. This is important for patients who use esophageal voice since voice samples mostly include type III (highly aperiodic) and 4 (chaotic) signals. As such, using non-linear measures, which are better suited for aperiodic phonation, may be useful to investigate the relationship between acoustic signal characteristics and perception of esophageal voice quality. This study aimed to investigate whether nonlinear dynamic acoustic methods, nonlinear energy difference Ratio (NEDR) and spectrum convergence ratio (SCR), were correlated with perceptual measures in subjects who used esophageal phonation. Thirty-one subjects who had undergone total laryngectomy and use esophageal voice as a rehabilitation method were included in this study. Expert and non-expert raters listened to the esophageal voice samples from the subjects and rated vowels and connected speech samples on a scale from 1 to 7 on dysphonia severity and intelligibility. In addition, non-linear acoustic analysis was performed to calculate NEDR and SCR. Analysis from the raters was compared to the non-linear acoustic analysis to find the correlation between the variables. There were no significant correlations between any of the non-linear acoustic measures NEDR and SCR and the perceptual ratings at the significance level of 0.05. Correlations were calculated for each acoustic measure among the expert raters and among the non-expert raters in both connected speech samples and sustained vowel fragments. In conclusion, the nonlinear dynamic acoustic analyses of spectrum convergence ratio and nonlinear energy difference ratio do not have a significant correlation with perceptual measures of esophageal voice.

The Effect of Different Preparatory Conducting Gestures on Breathing Behavior and Voice Quality of Choral Singers

The breathing technique is a determining factor for the singer's sound quality and consequently crucial for the choral sound. However, very little is known about possible influences of the conductor's preparatory gesture on the way choral singers inhale before the beginning of a piece (respectively every subsequent phrase). The conducting literature does not discriminate between out- and inward preparatory gestures and even describes them as equivalent, but previous studies suggest that singers assign different types of inhalation to different preparatory gestures. It may therefore be assumed that the type of preparatory gesture has a direct influence on the singer's inhalation and tone production, and the aim of this study is hence to examine possible effects of two contrasting preparatory gestures on the singer's inhalation type and the resulting tone quality. In our within-subjects study design, 18 healthy choral singers (9 male/ 9 female) were recruited to participate in a laboratory experiment. The participants were asked to sing a tone suitable for their voice register in response to different video stimuli. These consisted of two conducting-videos, each showing a different preparatory gesture, and two control conditions with an animated bar and an arrow indicating the desired breathing type. The singers reacted to 10 sets of videos, each set consisting of the four stimuli in randomized order. For evaluation of the breathing behavior and vocal output during the different experimental conditions, chest wall kinematics of upper rib cage, abdominal rib cage and abdomen were measured via 3D motion capture and voice samples were recorded. The obtained data were filtered and compared using the repeated measures analysis of variance and post hoc Tukey test for significant results. The level of significance was set at P < 0.05. The results of the study show significant differences in volume of the abdomen between the two different gestures (F1,17=24.04, η2=0.59, P=0.0001), which can be validated by the two control measurements (F1,17=21.12, η2=0.55, P=0.0002). An outward preparatory gesture evoked an abdominal breathing type while an inward-upward movement led to an inhalation with a higher portion of clavicular breathing. Furthermore, significant differences in timbre and loudness of the produced tone could be observed. The maximum sound pressure level of the outward preparatory gesture was significantly higher than in case of the inward-upward movement (F1,17=20.4, η2=0.56, P=0.0004). In contrast to the existing conducting literature, which does not discriminate between out- and inward preparatory gestures, the results of this study show that the conductor's choice of trajectory direction and form of the preparatory gestures elicit spontaneous, gesture-specific reactions in singers' breathing behavior as well as the corresponding loudness and sound quality.

A Randomized Controlled Trial With Female Teachers: Are there Differences Between and Within the Outcomes in Voice Therapy Groups With and Without Carryover Strategies?

We investigated if outcomes differ between voice therapy groups systematically using carryover strategies (attempts to generalize new vocal skills outside the clinic) and voice therapy with no emphasis on any generalizing process (here referred to as traditional voice therapy). A randomized controlled trial was conducted. Participants (53 female teachers with voice disorders) were randomly allocated into three groups: Carryover (a group receiving voice therapy using carryover strategies), Trad (a group receiving voice therapy with no emphasis on any generalizing process), Controls (a group on an eight-week non-therapy period). Prior to the trial a direct laryngoscopy was performed with a videolaryngostroboscopy system and/or nasofaryngofiberoscope with stroboscopy. Before and after therapy and at follow-up a voice evaluation protocol was implemented consisting of subjective assessments (Questionnaire on Voice Symptoms, and the Voice Activity and Participation Profile; VAPP), and objective measurements (voice sample recordings, acoustic analysis [SPL, sound pressure level; f0, fundamental frequency; alpha-ratio, tilt of the sound spectrum slope]). No differences were found between the groups. Several significant changes occurred within the groups between initial phase vs. post-therapy and initial phase vs. follow-up. In the Carryover group text reading the alpha-ratio became lower (P=0.011) and spontaneous speech f0 increased (P=0.024) after the therapy and [a:] SPL increased (P=0.042) at follow-up. In the Trad group post-therapy [a:] alpha-ratio became lower (P=0.012) and spontaneous speech f0 decreased (P=0.034). After therapy VAPP scores showed improvement in voice-related quality of life in both therapy groups (Carryover P=0.003; Trad P = 0.01) but only in Carryover at follow-up (P=0.000). Voice symptoms decreased in the Carryover group post-therapy (P=0.001) and at follow-up (P=0.000) and after Controls' eight-week non-therapy period (P=0.003). The results showed that carryover strategies give no additional advantages in voice therapy. However, the decreasing trend in the Carryover group's voice complaints at follow-up would suggest that carryover strategies may have long-lasting effects. The results also confirm that voice therapy is efficient in improving voice-related quality of life.

Gender Voice Recognition Using Machine Learning Algorithms

Abstract: Gender identification is one of the major problems in the area of signal processing. The system deals with finding the gender of a person using oral features. One of the most persnickety problems faced is feature selection from wide range of features, which is distinguishing factor in classifying the gender of aperson. The ideal of this design is to design a system that determines the speaker gender using the pitch of the speaker's voice. relating the gender from the plots of voice data set i.e., pitch, median, frequency etc. can be possible by using machine learning. In this design, we're trying to classify gender into male or female based on the data set containing varied attributes related to voice like pitch, frequency etc. The data set have features with explanation data points recorded samples of male and female voices. The data set can be trained with different machine learning algorithms. The proposed system can determine the gender of the speaker with real time test data a new result to discover the gender of the speaker using Fast Fourier Transform with Logistic Regression

Resynthesis of Transmasculine Voices to Assess Gender Perception as a Function of Testosterone Therapy.

The goal of this study was to use speech resynthesis to investigate the effects of changes to individual acoustic features on speech-based gender perception of transmasculine voice samples following the onset of hormone replacement therapy (HRT) with exogenous testosterone. We hypothesized that mean fundamental frequency (f o) would have the largest effect on gender perception of any single acoustic feature. Mean f o, f o contour, and formant frequencies were calculated for three pairs of transmasculine speech samples before and after HRT onset. Sixteen speech samples with unique combinations of these acoustic features from each pair of speech samples were resynthesized. Twenty young adult listeners evaluated each synthesized speech sample for gender perception and synthetic quality. Two analyses of variance were used to investigate the effects of acoustic features on gender perception and synthetic quality. Of the three acoustic features, mean f o was the only single feature that had a statistically significant effect on gender perception. Differences between the speech samples before and after HRT onset that were not captured by changes in f o and formant frequencies also had a statistically significant effect on gender perception. In these transmasculine voice samples, mean f o was the most important acoustic feature for voice masculinization as a result of HRT; future investigations in a larger number of transmasculine speakers and on the effects of behavioral therapy-based changes in concert with HRT is warranted.

Exploring Longitudinal Cough, Breath, and Voice Data for COVID-19 Progression Prediction via Sequential Deep Learning: Model Development and Validation.

BackgroundRecent work has shown the potential of using audio data (eg, cough, breathing, and voice) in the screening for COVID-19. However, these approaches only focus on one-off detection and detect the infection, given the current audio sample, but do not monitor disease progression in COVID-19. Limited exploration has been put forward to continuously monitor COVID-19 progression, especially recovery, through longitudinal audio data. Tracking disease progression characteristics and patterns of recovery could bring insights and lead to more timely treatment or treatment adjustment, as well as better resource management in health care systems.ObjectiveThe primary objective of this study is to explore the potential of longitudinal audio samples over time for COVID-19 progression prediction and, especially, recovery trend prediction using sequential deep learning techniques.MethodsCrowdsourced respiratory audio data, including breathing, cough, and voice samples, from 212 individuals over 5-385 days were analyzed, alongside their self-reported COVID-19 test results. We developed and validated a deep learning–enabled tracking tool using gated recurrent units (GRUs) to detect COVID-19 progression by exploring the audio dynamics of the individuals’ historical audio biomarkers. The investigation comprised 2 parts: (1) COVID-19 detection in terms of positive and negative (healthy) tests using sequential audio signals, which was primarily assessed in terms of the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity, with 95% CIs, and (2) longitudinal disease progression prediction over time in terms of probability of positive tests, which was evaluated using the correlation between the predicted probability trajectory and self-reported labels.ResultsWe first explored the benefits of capturing longitudinal dynamics of audio biomarkers for COVID-19 detection. The strong performance, yielding an AUROC of 0.79, a sensitivity of 0.75, and a specificity of 0.71 supported the effectiveness of the approach compared to methods that do not leverage longitudinal dynamics. We further examined the predicted disease progression trajectory, which displayed high consistency with longitudinal test results with a correlation of 0.75 in the test cohort and 0.86 in a subset of the test cohort with 12 (57.1%) of 21 COVID-19–positive participants who reported disease recovery. Our findings suggest that monitoring COVID-19 evolution via longitudinal audio data has potential in the tracking of individuals’ disease progression and recovery.ConclusionsAn audio-based COVID-19 progression monitoring system was developed using deep learning techniques, with strong performance showing high consistency between the predicted trajectory and the test results over time, especially for recovery trend predictions. This has good potential in the postpeak and postpandemic era that can help guide medical treatment and optimize hospital resource allocations. The changes in longitudinal audio samples, referred to as audio dynamics, are associated with COVID-19 progression; thus, modeling the audio dynamics can potentially capture the underlying disease progression process and further aid COVID-19 progression prediction. This framework provides a flexible, affordable, and timely tool for COVID-19 tracking, and more importantly, it also provides a proof of concept of how telemonitoring could be applicable to respiratory diseases monitoring, in general.

Validation of Acoustic Breathiness Index (ABI) in the South Indian Population

Acoustic Breathiness Index (ABI) is an acoustic model which equates to the auditory-perceptual (A-P) measure of breathiness. Globally, ABI has been validated in multiple languages. To date, empirical studies validating ABI in the South Indian population have not been carried out. Moreover, there is only minimal evidence on the impact of language and gender on ABI. This study attempted to validate ABI in languages of Malayalam and Kannada with "B" and "G" of GRBAS. The authors also examined the influence of language and gender on ABI. The authors retrospectively analyzed a dataset consisting of 160 voice samples (91 normophonic and 69 dysphonic). Five speech-language pathologists perceptually evaluated these samples on the GRBAS scale. PRAAT software was used to calculate ABI for each voice sample. The authors also determined the effects of language and gender on ABI. Following this, ABI's concurrent validity and diagnostic accuracy were examined relating to "B" and "G" of the GRBAS scale. The judges reached a moderate agreement on the perceptual evaluation of voice quality. The authors observed neither language nor gender differences in ABI. The concurrent validity of ABI was strong (r=0.761) when correlated against "B". Further, evaluations revealed that an ABI value of >3.66 could distinguish between breathy and non-breathy voice samples with an area under the curve of 0.816 (Sensitivity=61.73%; Specificity=94.93%). On the other hand, the concurrent validity of ABI in correlation with "G" did not yield the optimal result. ABI is a valid acoustic tool for identifying breathiness in the languages of Malayalam and Kannada. ABI correlates better with "B" than "G" on the GRBAS scale. ABI is independent of language and gender.

Cross-cultural adaptation of the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V) into Malay: A validity study

Auditory-perceptual rating of voice is one component of voice evaluation, and the CAPE-V is one of the tools for this purpose. Because of its advantages, the CAPE-V has been adapted into several languages. Accordingly, the adaptation of the CAPE-V into Malay is essential for its utility among the Malaysian population, which this study aimed to accomplish. This study involved translating the CAPE-V into Malay, termed the Malay CAPE-V, followed by establishing its validity. The translation processes (i.e., forward translation and backward translation) involved four different translators i.e., 3 speech-language therapists (SLTs) and 1 linguist with at least 14 years of experience. Most items were similarly translated, except for a few, which were subsequently accepted as similar, following consensus among the translators. In the examination of content validity, two raters (i.e., SLTs) with at least 12 years of clinical experience reviewed and scored the tested items. Given that the scale-level content validity index average value (S-CVI/Ave) was above the acceptable level, all items were retained. The examination of construct validity and concurrent validity involved ratings of voice samples recorded from 38 participants, comprising 19 individuals with normal and disordered voices, respectively. The ratings were performed by three raters (i.e., SLTs) with at least 12 years of clinical experience. For the construct validity, the Mann-Whitney U test indicated significantly higher scores of the vocal parameters of the Malay CAPE-V for the disordered voice group than for the normal voice group. Meanwhile, for the concurrent validity, the Spearman correlation indicated that all relationships between the scores of the vocal parameters of the Malay CAPE-V and GRBAS Scale were significant, with most of them achieving a very strong positive correlation. The results demonstrated that the Malay CAPE-V is a valid tool for an auditory-perceptual rating of voice among the Malaysian population.

Voice Isolation Using Artificial Neural Network

Abstract: The paper reflects the use of Artificial Neural Networks with the help of various machine learning algorithms for voice isolation. In particular, we consider the case of a voice sample recognition by analyzing the speech signals with the help of machine learning algorithms such as artificial neural networks, independent component analysis, activation function. The strategies by which our central nervous network decodes the network stimuli same as artificial neural network will analyze the given speech sample. After first step, a set of machine learning algorithms will be used like independent component analysis algorithm and gradient function algorithm for processing. After the processing, a decision statement will be applied to generate the desired output. Keywords: Artificial Neural Network, Voice Isolation, Fast Independent Component Analysis, Gradient Descent