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

Conducting high-quality and reliable acoustic analysis: A tutorial focused on training research assistants

Open science practices have led to an increase in available speech datasets for researchers interested in acoustic analysis. Accurate evaluation of these databases frequently requires manual or semi-automated analysis. The time-intensive nature of these analyses makes them ideally suited for research assistants in laboratories focused on speech and voice production. However, the completion of high-quality, consistent, and reliable analyses requires clear rules and guidelines for all research assistants to follow. This tutorial will provide information on training and mentoring research assistants to complete these analyses, covering areas including RA training, ongoing data analysis monitoring, and documentation needed for reliable and re-creatable findings.

Evaluating the Effect of Voice Quality Covariance on Auditory-Perceptual Evaluation Using a Novel Two-Dimensional Magnitude Estimation Task

PURPOSE

Most people with dysphonia present with voices that vary along more than one voice quality (VQ) dimension. This study sought to examine the effect of covariance between breathy and rough VQ in natural voices.

METHOD

A two-dimensional matrix of 16 /a/ vowels was selected such that two VQ dimensions (breathiness and roughness) were sampled on a 4-point severity scale (none, mild, moderate, and severe). Ten listeners evaluated 480 stimuli (16 stimuli × 10 repetitions × 3 blocks) on one-dimensional magnitude estimation (1DME) tasks and a novel two-dimensional magnitude estimation (2DME) task that allowed for simultaneous measurement of breathiness and roughness.

RESULTS

Data indicated high intra- and interrater reliabilities for both breathiness and roughness in the 2DME and 1DME tasks. Correlation analyses revealed a strong correlation between 2DME and 1DME judgments for breathiness and roughness (r > .95). There was also a minimal correlation between breathy and rough VQ in the 2DME task (r < .10).

CONCLUSIONS

Covarying roughness or breathiness had less impact on the perception of the other VQ in natural dysphonic voices in 2DME compared to 1DME. An understanding and quantification of the perceptual interactions among the dimensions will aid in the refinement of computational models and in the establishment of the validity of clinical scales for VQ perception.

Comparative Analysis of Two Methods of Perceptual Voice Assessment

OBJECTIVES

The primary aim was to compare two methods for perceptual evaluation of voice - paired comparison (PC) and visual analog scale (VAS) ratings. Secondary aims were to assess the correspondence between two dimensions of voice- overall severity of voice quality and resonant voice, and to investigate the influence of rater experience on perceptual rating scores and rating confidence scores.

STUDY DESIGN

Experimental design.

METHODS

Voice samples from six children (pre and post therapy) were rated by 15 Speech-Language Pathologists specialized in voice. Raters completed four tasks corresponding to the two rating methods and voice qualities: PC-severity, PC-resonance, VAS-severity, and VAS-resonance. For PC tasks, raters chose the better of two voice samples (better voice quality or better resonance, depending on the task) and indicated the degree of confidence in each choice. Rating and confidence score were combined to produce a number on a 1-10 scale (PC-confidence adjusted). VAS ratings involved rating voices on a scale for degree of severity and resonance, respectively.

RESULTS

PC-confidence adjusted and VAS ratings were moderately correlated for overall severity and also vocal resonance. VAS ratings were normally distributed and had greater rater consistency than PC-confidence adjusted ratings. VAS scores reliably predicted binary PC choices (choice of voice sample only). Overall severity and vocal resonance were weakly correlated and rater experience was not linearly related to rating scores or confidence.

CONCLUSIONS

Results suggest that the VAS rating method holds advantages over PC, including normally distributed ratings, superior consistency of ratings, and the ability to provide more finely grained detail regarding the auditory perception of voice. Overall severity and vocal resonance were not redundant in the current data set, suggesting that resonant voice and overall severity are not isomorphic. Finally, the number of years of clinical experience was not linearly related to perceptual ratings or rating confidence.