The intensive use of personal protective equipment often requires increasing voice intensity, with possible development of voice disorders. This paper exploits machine learning approaches to investigate the impact of different types of masks on sustained vowels /a/, /i/, and /u/ and the sequence /a'jw/ inside a standardized sentence. Both objective acoustical parameters and subjective ratings were used for statistical analysis, multiple comparisons, and in multivariate machine learning classification experiments. Significant differences were found between mask+shield configuration and no-mask and between mask and mask+shield conditions. Power spectral density decreases with statistical significance above 1.5 kHz when wearing masks. Subjective ratings confirmed increasing discomfort from no-mask condition to protective masks and shield. Machine learning techniques proved that masks alter voice production: in a multiclass experiment, random forest (RF) models were able to distinguish amongst seven masks conditions with up to 94% validation accuracy, separating masked from unmasked conditions with up to 100% validation accuracy and detecting the shield presence with up to 86% validation accuracy. Moreover, an RF classifier allowed distinguishing male from female subject in masked conditions with 100% validation accuracy. Combining acoustic and perceptual analysis represents a robust approach to characterize masks configurations and quantify the corresponding level of discomfort.

No abstract

Adductor spasmodic dysphonia is a type of adult-onset focal dystonia characterized by involuntary spasms of laryngeal muscles. This paper applied machine learning techniques for the severity assessment of spasmodic dysphonia. To this aim, 7 perceptual indices and 48 acoustical parameters were estimated from the Italian word /a’jwɔle/ emitted by 28 female patients, manually segmented from a standardized sentence and used as features in two classification experiments. Subjects were divided into three severity classes (mild, moderate, severe) on the basis of the G (grade) score of the GRB scale. The first aim was that of finding relationships between perceptual and objective measures with the Local Interpretable Model-Agnostic Explanations method. Then, the development of a diagnostic tool for adductor spasmodic dysphonia severity assessment was investigated. Reliable relationships between G; R (Roughness); B (Breathiness); Spasmodicity; and the acoustical parameters: voiced percentage, F2 median, and F1 median were found. After data scaling, Bayesian hyperparameter optimization, and leave-one-out cross-validation, a k-nearest neighbors model provided 89% accuracy in distinguishing patients among the three severity classes. The proposed methods highlighted the best acoustical parameters that could be used jointly with GRB indices to support the perceptual evaluation of spasmodic dysphonia and provide a tool to help severity assessment of spasmodic dysphonia.