Measuring vocal quality with speech synthesis.

Much previous research has demonstrated that listeners do not agree well when using traditional rating scales to measure pathological voice quality. Although these findings may indicate that listeners are inherently unable to agree in their perception of such complex auditory stimuli, another explanation implicates the particular measurement method-rating scale judgments-as the culprit. An alternative method of assessing quality-listener-mediated analysis-synthesis-was devised to assess this possibility. In this new approach, listeners explicitly compare synthetic and natural voice samples, and adjust speech synthesizer parameters to create auditory matches to voice stimuli. This method is designed to replace unstable internal standards for qualities like breathiness and roughness with externally presented stimuli, to overcome major hypothetical sources of disagreement in rating scale judgments. In a preliminary test of the reliability of this method, listeners were asked to adjust the signal-to-noise ratio for 12 synthetic pathological voices so that the resulting stimuli matched the natural target voices as well as possible For comparison to the synthesis judgments, listeners also judged the noisiness of the natural stimuli in a separate task using a traditional visual-analog rating scale. For 9 of the 12 voices, agreement among listeners was significantly (and substantially) greater for the synthesis task than for the rating scale task. Response variances for the two tasks did not differ for the remaining three voices. However, a second experiment showed that the synthesis settings that listeners selected for these three voices were within a difference limen, and therefore observed differences were perceptually insignificant. These results indicate that listeners can in fact agree in their perceptual assessments of voice quality, and that analysis-synthesis can measure perception reliably.

Sources of listener disagreement in voice quality assessment.

Traditional interval or ordinal rating scale protocols appear to be poorly suited to measuring vocal quality. To investigate why this might be so, listeners were asked to classify pathological voices as having or not having different voice qualities. It was reasoned that this simple task would allow listeners to focus on the kind of quality a voice had, rather than how much of a quality it possessed, and thus might provide evidence for the validity of traditional vocal qualities. In experiment 1, listeners judged whether natural pathological voice samples were or were not primarily breathy and rough. Listener agreement in both tasks was above chance, but listeners agreed poorly that individual voices belonged in particular perceptual classes. To determine whether these results reflect listeners' difficulty agreeing about single perceptual attributes of complex stimuli, listeners in experiment 2 classified natural pathological voices and synthetic stimuli (varying in f0 only) as low pitched or not low pitched. If disagreements derive from difficulties dividing an auditory continuum consistently, then patterns of agreement should be similar for both kinds of stimuli. In fact, listener agreement was significantly better for the synthetic stimuli than for the natural voices. Difficulty isolating single perceptual dimensions of complex stimuli thus appears to be one reason why traditional unidimensional rating protocols are unsuited to measuring pathologic voice quality. Listeners did agree that a few aphonic voices were breathy, and that a few voices with prominent vocal fry and/or interharmonics were rough. These few cases of agreement may have occurred because the acoustic characteristics of the voices in question corresponded to the limiting case of the quality being judged. Values of f0 that generated listener agreement in experiment 2 were more extreme for natural than for synthetic stimuli, consistent with this interpretation.

Combined arytenoid adduction and laryngeal reinnervation in the treatment of vocal fold paralysis.

OBJECTIVE/HYPOTHESIS: Glottal closure and symmetrical thyroarytenoid stiffness are two important functional characteristics of normal phonatory posture. In the treatment of unilateral vocal cord paralysis, vocal fold medialization improves closure, facilitating entrainment of both vocal folds for improved phonation, and reinnervation is purported to maintain vocal fold bulk and stiffness. A combination of medialization and reinnervation would be expected to further improve vocal quality over medialization alone. STUDY DESIGN: A retrospective review of preoperative and postoperative voice analysis on all patients who underwent arytenoid adduction alone (adduction group) or combined arytenoid adduction and ansa cervicalis to recurrent laryngeal nerve anastomosis (combined group) between 1989 and 1995 for the treatment of unilateral vocal cord paralysis. Patients without postoperative voice analysis were invited back for its completion. A perceptual analysis was designed and completed. METHODS: Videostroboscopic measures of glottal closure, mucosal wave, and symmetry were rated. Aerodynamic parameters of laryngeal airflow and subglottic pressure were measured. A 2-second segment of sustained vowel was used for perceptual analysis by means of a panel of voice professionals and a rating system. Statistical calculations were performed at a significance level of P = .05. RESULTS: There were 9 patients in the adduction group and 10 patients in the combined group. Closure and mucosal wave improved significantly in both groups. Airflow decreased in both groups, but the decrease reached statistical significance only in the adduction group. Subglottic pressure remained unchanged in both groups. Both groups had significant perceptual improvement of voice quality. In all tested parameters the extent of improvement was similar in both groups. CONCLUSION: The role of laryngeal reinnervation in the treatment of unilateral vocal cord paralysis remains to be established.

Exit jet particle velocity in the in vivo canine laryngeal model with variable nerve stimulation.

This study extends previous work on exit jet particle velocity in the in vivo canine model of phonation by measuring air particle velocity at multiple locations in the midline of the glottis and across multiple levels of recurrent laryngeal nerve (RLN) and superior laryngeal nerve (SLN) stimulation. In a second experiment, exit jet particle velocity was measured at midline and offmidline positions with constant levels of RLN and SLN stimulation. In this study, peak particle velocity was higher at the anterior commissure than at the posterior commissure in the midline of the glottis, and peak particle velocity was higher at the midline than at offmidline positions. In addition, increasing levels of RLN stimulation resulted in increasing peak particle velocity; however, increasing levels of SLN stimulation failed to produce a uniform effect on peak particle velocity.

Selective laryngeal adductor denervation-reinnervation: a new surgical treatment for adductor spasmodic dysphonia.

During the past decade, botulinum toxin (Botox) has emerged as the accepted treatment for adductor spasmodic dysphonia (ASD). This therapy, which produces bilateral weakness of the thyroarytenoid muscle, undoubtedly produces physiologic effects that are beneficial to patients with ASD. However, it also has important limitations, including the need for repeated injections, the unpredictable relationship between dosage and response, and the possibility of short-term swallowing and voice problems. In this study, we will report our preliminary experience with a new surgical treatment for ASD. In this new procedure, the adductor branch of the recurrent laryngeal nerve is selectively denervated bilaterally, and its distal nerve stumps are reinnervated with branches of the ansa cervicalis nerve. Each of the patients was followed for at least 12 months; the median follow-up is 36 months. The outcome of the operation in 21 consecutive patients is reported. Nineteen of the 21 patients were judged to have an overall severity of dysphonia that was "absent to mild" following the procedure. Only 1 patient underwent further treatment with Botox postoperatively. The implications of this new procedure for ASD are discussed.

Validity of rating scale measures of voice quality.

The validity of perceptual measures of vocal quality has been neglected in studies of voice, which focus more commonly on rater reliability. Validity depends in part on reliability, because an unreliable test does not measure what it is intended to measure. However, traditional measures of rating reliability only partially represent interrater agreement, because they cannot reflect variations or patterns of agreement for specific voice samples. In this paper the likelihood that two raters would agree in their ratings of a single voice is examined, for each voice in five previously gathered data sets. Results do not support the continued assumption that traditional rating procedures produce useful indices of listeners' perceptions. Listeners agreed very poorly in the midrange of scales for breathiness and roughness, and mean ratings in the midrange of such scales did not represent the extent to which a voice possesses a quality, but served only to indicate that listeners disagreed. Techniques like analysis by synthesis or judgment of similarity avoid decomposing quality into constituent dimensions, and do not require a listener to compare an external stimulus to an unstable internal representation, thus decreasing the error in measures of quality. Modeling individual differences in perception can increase the variance accounted for in models of quality, further reducing the error in perceptual measures. Thus such techniques may provide valid alternatives to current approaches.

The perceptual structure of pathologic voice quality.

Although perceptual assessment is included in most protocols for evaluating pathologic voices, a standard set of valid scales for measuring voice quality has never been established. Standardization is important for theory and for clinical acceptance, and also because validation of objective measures of voice depends on valid perceptual measures. The present study used large sets (n = 80) of male and female voices, representing a broad range of diagnoses and vocal severities. Eight experts judged the dissimilarity of each pair of voices, and responses were analyzed using nonmetric individual differences multidimensional scaling. Results indicate that differences between listeners in perceptual strategy are so great that the fundamental assumption of a common perceptual space must be questioned. Because standardization depends on the assumption that listeners are similar, it is concluded that efforts to standardize perceptual labels for voice quality are unlikely to succeed. However, analysis by synthesis may provide an alternate means of modeling quality as a function of both voices and listeners, thus avoiding this problem.

Effects of driving pressure and recurrent laryngeal nerve stimulation on glottic vibration in a constant pressure model.

Glottic phonatory parameters have been studied in constant flow models; however, the lung-thorax system is better viewed as a constant pressure source. Adjusting the driving pressure and recurrent laryngeal nerve stimulation as independent variables, rather than as dependent variables, may provide a more physiologic understanding of laryngeal function and glottic parameters, including subglottic pressure, airflow, fundamental frequency, and glottic area. In three dogs subglottic pressure and airflow were measured in two separate conditions: with constant recurrent laryngeal nerve stimulation and varying driving pressure, and with constant driving pressure and varying recurrent laryngeal nerve stimulation. Videostroboscopic measures on four dogs assessed glottic areas with constant recurrent laryngeal nerve stimulation at different driving pressures. With constant recurrent laryngeal nerve stimulation, increasing driving pressure had no effect on glottic areas, whereas subglottic pressure, fundamental frequency, and airflow increased significantly. However, changes in subglottic pressure were minimal in comparison with changes in driving pressure. At constant driving pressure, increasing recurrent laryngeal nerve stimulation increased subglottic pressure and fundamental frequency and decreased airflow. These findings suggest that during phonation subglottic pressure is primarily dependent on recurrent laryngeal nerve stimulation and laryngeal muscular contraction, but not on lung driving pressure.

Characteristics of an in vivo canine model of phonation with a constant air pressure source.

Many previous studies of laryngeal biomechanics using in vivo models have employed a constant air How source. Several authors have recently suggested that the lung-thorax system functions as a constant pressure source during phonation. This study describes an in vivo canine system designed to maintain a constant peak subglottic pressure (Psub) using a pressure-controlling mechanism. Increasing levels of recurrent laryngeal nerve (RLN) stimulation resulted in a significant rise in resistance followed by a plateau. For a given Psub, flow decreased significantly and precipitously with increasing stimulation and then quickly plateaued. Vocal intensity increased with increasing RLN stimulation until a peak was reached. After this peak, intensity dropped until a plateau was reached, corresponding to the flow minimum. At a given Psub, increasing levels of RLN stimulation resulted in a normal distribution of vocal efficiencies.

Comparison of voice analysis systems for perturbation measurement.

Dysphonic voices are often analyzed using automated voice analysis software. However, the reliability of acoustic measures obtained from these programs remains unknown, particularly when they are applied to pathological voices. This study compared perturbation measures from CSpeech, Computerized Speech Laboratory, SoundScope, and a hand marking voice analysis system. Sustained vowels from 29 male and 21 female speakers with mild to severe dysphonia were digitized, and fundamental frequency (F0), jitter, shimmer, and harmonics- or signal-to-noise ratios were computed. Commercially available acoustical analysis programs agreed well, but not perfectly, in their measures of F0. Measures of perturbation in the various analysis packages use different algorithms, provide results in different units, and often yield values for voices that violate the assumption of quasi-periodicity. As a result, poor rank order correlations between programs using similar measures of perturbation were noted. Because measures of aperiodicity apparently cannot be reliably applied to voices that are even mildly aperiodic, we question their utility in quantifying vocal quality, especially in pathological voices.

A comparison of type I thyroplasty and arytenoid adduction.

Glottal incompetence is a common laryngeal disorder causing impaired swallowing and phonation. The resultant voice has been characterized as weak and breathy with a restricted pitch range. Currently, medialization thyroplasty and arytenoid adduction are two of the surgical treatments for patients with glottal incompetence. However, few studies have evaluated the changes in objective measures of speech with type I thyroplasty and arytenoid adduction. In this study, 59 patients with glottal incompetence underwent either type I thyroplasty or arytenoid adduction. Acoustic (jitter, shimmer, and harmonics-to-noise ratio) and aerodynamic (airflow, subglottic pressure, and glottal resistance) measures were obtained both pre- and postoperatively. No significant differences were found among acoustic or aerodynamic measures for operation type. However, a significant pre/postsurgery effect was observed for translaryngeal airflow. In addition, no significant differences were found among the measures for patients with traditional compared with nontraditional operative indications. Patients who developed glottal insufficiency due to previous laryngeal surgery (e.g., vocal fold stripping) demonstrated no statistically significant improvement in acoustic or aerodynamic measures following thyroplasty or arytenoid adduction.

The effect of gas density on glottal vibration and exit jet particle velocity.

Although theoretical studies include a term for gas density in their mathematical descriptions of glottal aerodynamics, the effect of gas density on glottal vibration has not been examined empirically. In this study, an in vivo canine model was used to evaluate the effect of gas density on glottal vibration by comparing phonation with air and helium. With gas flow and nerve stimulation held constant, phonation with helium resulted in an increased exit jet particle velocity for helium (45 m/s) compared to air (34 m/s). However, the measured increase in helium velocity was less than predicted by a proportional relationship between transglottal pressure and dynamic pressure. This difference could be due to a change in the constant of proportionality or in the dynamic pressure loss coefficient associated with the use of helium.

Comparing reliability of perceptual ratings of roughness and acoustic measure of jitter.

Acoustic analysis is often favored over perceptual evaluation of voice because it is considered objective, and thus reliable. However, recent studies suggest this traditional bias is unwarranted. This study examined the relative reliability of human listeners and automatic systems for measuring perturbation in the evaluation of pathologic voices. Ten experienced listeners rated the roughness of 50 voice samples (ranging from normal to severely disordered) on a 75 mm visual analog scale. Rating reliability within and across listeners was compared to the reliability of jitter measures produced by several voice analysis systems (CSpeech, SoundScope, CSL, and an interactive hand-marking system). Results showed that overall listeners agreed as well or better than "objective" algorithms. Further, listeners disagreed in predictable ways, whereas automatic algorithms differed in seemingly random fashions. Finally, listener reliability increased with severity of pathology; objective methods quickly broke down as severity increased. These findings suggest that listeners and analysis packages differ greatly in their measurement characteristics. Acoustic measures may have advantages over perceptual measures for discriminating among essentially normal voices; however, reliability is not a good reason for preferring acoustic measures of perturbation to perceptual measures.

A pressure-regulated model of normal and pathologic phonation.

Recent evidence suggests that the lung-thorax system functions as a constant pressure source during phonation. However, previous animal models used a constant flow source. This article describes an in vivo canine model that maintains a constant subglottic pressure during phonation to more closely simulate the pulmonary system. At any given subglottic pressure, increasing levels of recurrent laryngeal nerve stimulation resulted in a significant rise in resistance followed by a plateau. Increasing levels of superior laryngeal nerve stimulation, however, produced no significant change in glottal resistance. Three experimental conditions were studied: normal, unilateral recurrent laryngeal nerve paralysis, and paralysis followed by arytenoid adduction. In normal canines, maximal vocal efficiency values were the highest, indicating the best match between pressure and resistance. The vocal efficiency values were significantly lower in recurrent laryngeal nerve paralysis, indicating pressure-resistance mis-match. Arytenoid adduction increased the maximal vocal efficiency values and decreased the mismatch observed in the paralyzed state. These findings may provide insight into an understanding of normal and pathologic laryngeal behavior.

Effects of arytenoid adduction on laryngeal function following ansa cervicalis nerve transfer for vocal fold paralysis in an in vivo canine model.

Laryngeal reinnervation with the ansa cervicalis has been proposed as a treatment for human unilateral vocal fold paralysis (UVFP). This study tested the assumption that results from reinnervation could be improved if combined with medialization surgery. Six canine subjects underwent recurrent laryngeal nerve section and reinnervation with a branch of the ansa cervicalis. After reinnervation, vocal function was assessed before and after arytenoid adduction. Although laryngeal function improved significantly following reinnervation, results were significantly enhanced by the addition of medialization surgery. The implications for the treatment of human unilateral vocal fold paralysis are discussed.

The multidimensional nature of pathologic vocal quality.

Although the terms "breathy" and "rough" are frequently applied to pathological voices, widely accepted definitions are not available and the relationship between these qualities is not understood. To investigate these matters, expert listeners judged the dissimilarity of pathological voices with respect to breathiness and roughness. A second group of listeners rated the voices on unidimensional scales for the same qualities. Multidimensional scaling analyses suggested that breathiness and roughness are related, multidimensional constructs. Unidimensional ratings of both breathiness and roughness were necessary to describe patterns of similarity with respect to either quality. Listeners differed in the relative importance given to different aspects of voice quality, particularly when judging roughness. The presence of roughness in a voice did not appear to influence raters' judgments of breathiness; however, judgments of roughness were heavily influenced by the degree of breathiness, the particular nature of the influence varying from listener to listener. Differences in how listeners focus their attention on the different aspects of multidimensional perceptual qualities apparently are a significant source of interrater unreliability (noise) in voice quality ratings.

Effects of RLN and SLN stimulation on glottal area.

In vivo canine experiments have demonstrated that vocal fold stiffness varies proportionately with changing levels of recurrent laryngeal nerve (RLN) and superior laryngeal nerve (SLN) stimulation. This study evaluated the morphologic changes in the glottis at varying levels of nerve stimulation and the presumed effects on laryngeal air particle velocity. Stroboscopic data from the in vivo canine model of phonation were examined under varying conditions of RLN and SLN stimulation. Computerized analysis of stroboscopic images was used to reconstruct the glottal area vs. time waveforms. As RLN stimulation increased, glottal area per cycle decreased (p < 0.05). However, as SLN stimulation increased, glottal area per cycle increased (p < 0.05). However, as SLN stimulation increased, glottal area per cycle increased (p < 0.05). These results support the hypothesis that increasing RLN stimulation at similar levels of SLN stimulation produces an increase in air particle velocity, whereas an increase in SLN stimulation causes a decrease in air particle velocity.

Effect of tension, stiffness, and airflow on laryngeal resistance in the in vivo canine model.

This study used an in vivo canine model of phonation to determine the effects of airflow on glottal resistance at low, medium, and high levels of recurrent laryngeal nerve (RLN) and superior laryngeal nerve (SLN) stimulation. Static and dynamic trials of changing airflow were used to study the effects of airflow on glottal resistance during phonation. As reported previously, glottal resistance varies inversely as a function of airflow. Increasing levels of RLN stimulation resulted in a statistically significant increase in glottal resistance for each level of airflow evaluated. Variation in SLN stimulation had no statistically effects on the relationship between flow and resistance. At airflow rates greater than 590 milliliters per second (mL/s), glottal resistance approached 0.1 mm Hg per mL/s for all levels of RLN and SLN stimulation tested. These data support the collapsible tube model of phonation.