Marc De Bodt

Toward Improved Ecological Validity in the Acoustic Measurement of Overall Voice Quality: Combining Continuous Speech and Sustained Vowels

To improve ecological validity, perceptual and instrumental assessment of disordered voice, including overall voice quality, should ideally sample both sustained vowels and continuous speech. This investigation assessed the utility of combining both voice contexts for the purpose of auditory-perceptual ratings as well as acoustic measurement of overall voice quality. Sustained vowel and continuous speech samples from 251 subjects with (n=229) or without (n=22) various voice disorders were concatenated and perceptually rated on overall voice quality by five experienced voice clinicians. After removing the nonvoiced segments within the continuous speech samples, the concatenated samples were analyzed using 13 acoustic measures based on fundamental frequency perturbation, amplitude perturbation, spectral and cepstral analyses. Stepwise multiple regression analysis yielded a six-variable acoustic model for the multiparametric measurement of overall voice quality of the concatenated samples (with a cepstral measure as the main contributor to the prediction of overall voice quality). The correlation of this model with mean ratings of overall voice quality resulted in r(s)=0.78. A cross-validation approach involving the iterated internal cross-correlations with 30 subgroups of 100, 50, and 10 samples confirmed a comparable degree of association. Furthermore, the ability of the model to distinguish voice-disordered from vocally normal participants was assessed using estimates of diagnostic precision including receiver operating characteristic (ROC) curve analysis, sensitivity, and specificity, as well as likelihood ratios (LRs), which adjust for base-rate differences between the groups. Depending on the cutoff criteria employed, the analyses revealed an impressive area under ROC=0.895 as well as respectable sensitivity, specificity, and LR. The results support the diagnostic utility of combining voice samples from both continuous speech and sustained vowels in acoustic and perceptual analysis of disordered voice. The findings are discussed in relation to the extant literature and the need for further refinement of the acoustic algorithm.

Is the reliability of a visual analog scale higher than an ordinal scale? An experiment with the GRBAS scale for the perceptual evaluation of dysphonia.

Perceptual evaluation of 14 pathological voices was performed by 29 listeners using the GRBAS scale. To evaluate scale effects on the judgments, 2 versions of the scale were presented: the original 4-point scale and a visual analog scale. Each listener used the same voice samples for both versions of the scale with an interval of 2 weeks. Agreement was found to be higher with the original 4-point scale than with the visual analog version for all scale items G, R, B, A and S. Although a visual analog scale seems to enable a finer judgment of voice quality, this study showed that, with increased freedom of judgment, the interrater agreement decreased considerably. Therefore, we recommend the use of the original 4-point version of the GRBAS scale.

Acoustic measurement of overall voice quality: A meta-analysisa)

Over the past several decades, many acoustic markers have been proposed to be sensitive to and measure overall voice quality. This meta-analysis presents a retrospective appraisal of scientific reports, which evaluated the relation between perceived overall voice quality and several acoustic-phonetic correlates. Twenty-five studies met the inclusion criteria and were evaluated using meta-analytic techniques. Correlation coefficients between perceptual judgments and acoustic measures were computed. Where more than one correlation coefficient for a specific acoustic marker was available, a weighted average correlation coefficient was calculated. This was the case in 36 acoustic measures on sustained vowels and in 3 measures on continuous speech. Acoustic measures were ranked according to the strength of the correlation with perceptual voice quality ratings. Acoustic markers with more than one correlation value available in literature and yielding a homogeneous weighted r of 0.60 or above were considered to be superior. The meta-analysis identified four measures that met these criteria in sustained vowels and three measures in continuous speech. Although acoustic measures are routinely utilized in clinical voice examinations, the results of this meta-analysis suggest that caution is warranted regarding the concurrent validity and thus the clinical utility of many of these measures.

Intelligibility as a linear combination of dimensions in dysarthric speech.

Abstract Speech samples of 79 dysarthric patients (native speakers of English) were judged by two experienced judges by means of a perceptual rating scale covering the four main dimensions of speech production: voice quality, articulation, nasality and prosody as well as overall intelligibility. In order to determine the relative influence of the four basic dimensions on intelligibility in this study group, a multiple regression model was applied. This model shows that intelligibility can be expressed as a linear combination of weighted perceptual dimensions ( R 2 =0.89). This expression reveals the relative importance of the individual dimensions on overall intelligibility. Articulation is shown to be the strongest contributor to intelligibility. A group of 16 dysarthric patients (native speakers of Dutch) was used as a control group in order to evaluate the validity of the expression. Judged and calculated ratings (within the 95% prediction intervals) were in agreement for 75% of the patients. These findings may contribute in the construction of a better balanced measure of functional intelligibility. Learning outcomes 1. Intelligibility can be expressed by a linear combination of weighted single dimensions indicating the relative impact of each of them. 2. The impact of articulation on intelligibility is dominant but inclusion of the dimensions “nasality”, “voice” and “prosody” results in a more balanced estimation of intelligibility.

Evaluation of the Vocal Performance of Children Using a Voice Range Profile Index

Voice range profiles (VRPs) were analyzed according to 11 frequency, intensity, and morphological characteristics for 94 normal children and 136 children with vocal fold pathologies (ages 6-11 years). Normative data are presented showing marked differences between the groups. Using a specific combination of the childs age, the highest vocal fundamental frequency, the lowest intensity, and the slope of the upper VRP contour, a Voice Range Profile Index for Children (VRPIc) may be constructed using discriminant analysis. It is shown how the VRPIc can be used to screen children for vocal disorder or to quantitatively assess the effectiveness of voice treatment. Since the group means of the VRPIc for healthy and dysphonic children are scaled to +10 and -10, respectively, the VRPIc enables the clinician to rate a childs vocal performance with reference to healthy and dysphonic children in general. The sensitivity and specificity of this method was found to be 90% and 83%, respectively.

The Acoustic Voice Quality Index: Toward improved treatment outcomes assessment in voice disorders

UNLABELLED Voice practitioners require an objective index of dysphonia severity as a means to reliably track treatment outcomes. To ensure ecological validity however, such a measure should survey both sustained vowels and continuous speech. In an earlier study, a multivariate acoustic model referred to as the Acoustic Voice Quality Index (AVQI), consisting of a weighted combination of 6 time-, frequency- and quefrency-domain metrics, was developed to measure dysphonia severity in both speaking tasks. In the current investigation, the generalizability and clinical utility of the AVQI are evaluated by first assessing its external cross-validity and then determining its sensitivity to change in dysphonia severity following surgical and/or behavioral voice treatment. The results, based upon a new set of normal and disordered voices compared favorably with outcomes reported earlier, indicating acceptable external validity. Furthermore, the AVQI was sensitive to treatment-related changes, validating its role as a potentially robust and objective voice treatment outcomes measure. LEARNING OUTCOMES Readers will be able to: (1) explain methodological issues surrounding the development of voice treatment outcomes measures (such as external cross-validity and responsiveness to change), (2) appreciate the relevance of measuring dysphonia severity in both sustained vowels and connected speech, (3) describe the method of obtaining the Acoustic Voice Quality Index (AVQI), (4) appreciate differences among a variety of estimates of diagnostic accuracy, and (5) discuss the AVQI as a clinically valid treatment outcomes measure.

Training outcome in future professional voice users after 18 months of voice training.

The goal of this study is to define the long-term influence of vocal hygiene education and the effectiveness of voice training in 46 students. Half of the subjects, called the trained group (n = 23), received vocal hygiene education during 1 school year and voice training during 2 school years (18 months). The other half, also 23 subjects, received neither vocal hygiene education nor voice training as such (called the untrained group). The voice training is made up of technical workshops (30 h a year in groups of 5–8 subjects) and vocal coaching in the radio and drama projects (30 h whole class). In the lectures (30 h) a theoretical background on breathing, articulation, voicing and vocal hygiene was discussed. A multidimensional test battery containing the GRBAS scale, videolaryngostroboscopy, maximum phonation time, jitter, lowest intensity, highest frequency, Dysphonia Severity Index (DSI) and Voice Handicap Index (VHI) was applied before and after 18 months to evaluate the effect of voice training over time. A questionnaire on daily habits was presented before the lectures, and after 18 months to detect the long-term effect of the lectures. The objectively measured voice quality (DSI) of the trained group improved significantly over time (p < 0.001) due to training (p = 0.008), which was not the case in the untrained group. The self-assessed VHI, on the other hand, changed over time (p < 0.001) in both groups. For the trained group the VHI changed from 18.4 to 14.4 and in the untrained group from 20.1 to 15.3. It is important to note that the VHI scores of both groups remained high. The interpretation of the results of the daily habit questionnaire is disturbing: the initial high degree of smoking, vocal abuse, stress and late meals was not influenced by the lectures or training and remained high. This study proves the positive outcome and emphasizes the need for a well-organized voice training program in future professional voice users. However, the lectures and training on vocal hygiene failed to improve voice-conserving habits.

Cochlear implantation as a durable tinnitus treatment in single-sided deafness

Abstract Introduction Severe tinnitus can seriously impair patients in their activities in daily life and reduce their quality of life. The aims of this prospective clinical study were to assess the long-term effects of cochlear implantation (CI) on tinnitus in patients with single-sided deafness and ipsilateral incapacitating tinnitus, and to investigate whether CI could treat various types of tinnitus. Materials and methods Twenty-six subjects with unilateral severe-to-profound sensorineural hearing loss received a CI. Patients suffered from severe tinnitus greater than 6/10 on a visual analogue scale (VAS) due to unilateral deafness. Assessment consisted of a tinnitus analysis including determination of tinnitus type, frequency, and loudness. A tinnitus questionnaire (TQ) measured tinnitus distress. VAS and TQ were administered pre-implantation and post-implantation. Results All 26 patients reported a subjective benefit from CI. Tinnitus loudness reduced significantly after CI from 8.6 to 2.2 on the VAS (scale: 0–10). The TQ total score decreased significantly and the mean tinnitus degree decreased from severe to mild. No differences were observed between patients with pure-tone tinnitus, narrow band noise tinnitus, or polyphonic tinnitus. The degree of tinnitus loudness reduction remained stable after CI. Conclusions CI can successfully be used as treatment of severe tinnitus in patients with single-sided deafness and is equally effective for pure tone, narrow band noise, and polyphonic tinnitus. Long-term results show that implantation provides durable tinnitus relief in these patients. These results support the hypothesis that physiopathological mechanisms after peripheral deafferentation are reversible when hearing is restored. Single-sided deafness accompanied by severe tinnitus is a new indication for CI.

Automated intelligibility assessment of pathological speech using phonological features

It is commonly acknowledged that word or phoneme intelligibility is an important criterion in the assessment of the communication efficiency of a pathological speaker. People have therefore put a lot of effort in the design of perceptual intelligibility rating tests. These tests usually have the drawback that they employ unnatural speech material (e.g., nonsense words) and that they cannot fully exclude errors due to listener bias. Therefore, there is a growing interest in the application of objective automatic speech recognition technology to automate the intelligibility assessment. Current research is headed towards the design of automated methods which can be shown to produce ratings that correspond well with those emerging from a well-designed and well-performed perceptual test. In this paper, a novel methodology that is built on previous work (Middag et al., 2008) is presented. It utilizes phonological features, automatic speech alignment based on acoustic models that were trained on normal speech, context-dependent speaker feature extraction, and intelligibility prediction based on a small model that can be trained on pathological speech samples. The experimental evaluation of the new system reveals that the root mean squared error of the discrepancies between perceived and computed intelligibilities can be as low as 8 on a scale of 0 to 100.

Perturbation Measures of Voice: A Comparative Study between Multi-Dimensional Voice Program and Praat

Background/Aims: Frequency and amplitude perturbations are inherent in voice acoustic signals. The assessment of voice perturbation is influenced by several factors, including the type of recording equipment used and the measurement extraction algorithm applied. In the present study, perturbation measures provided by two computer systems (a purpose-built professional voice analysis apparatus and a personal computer-based system for acoustic voice assessment) and two computer programs (Multi-Dimensional Voice Program and Praat) were compared. Methods: Correlations and inferential statistics for seven perturbation measures (absolute jitter, percent jitter, relative average perturbation, pitch perturbation quotient, shimmer in decibels, percent shimmer, and amplitude perturbation quotient) in 50 subjects with various voice disorders are presented. Results: Results indicate statistically significant differences between the two systems and programs, with the Multi-Dimensional Voice Program yielding consistently higher measures than Praat. Furthermore, correlation analyses show weak to moderate proportional relationships between the two systems and weak to strong proportional relationships between the two programs. Conclusion: Based on the literature and the proportional relationships and differences between the two systems and programs under consideration in this study, one can state that one can hardly compare frequency perturbation outcomes across systems and programs and amplitude perturbation outcomes across systems.