Effects of vocal training on quantitative voice parameters in healthy voice adults and children

Objective. To assess and to compare healthy trained and untrained voices in adults and children by combined voice range profile and aerodynamic tests, to define the normal range of quantitative voice parameters and to select the most informative of them for discrimination between healthy trained and untrained voices. Material and Methods. 207 adults and 142 children before mutation with healthy voices were investigated. According to voice training subjects were divided in to the separate groups: untrained voice groups consisted of 127 untrained voice adults (38 males, 89 females) and 61 children (36 boys, 25 girls) and professionally trained voice groups consisted of 80 trained voice adults (21 males, 59 females) and 81 children (50 boys, 31 girls). Eighteen quantitative voice parameters were analysed: eight of voice range profile (VRP), 8 of speech range profile, coefficient of sound of spectral voice range profile, and aerodynamic maximum phonation time. Results. Analysis showed that healthy vocally trained adults and children demonstrate expanded vocal capabilities in comparison to healthy untrained voice subjects. Mean values of quantitative VRP parameters: pitch range, high frequency limit, area of high frequencies differ significantly between healthy trained versus untrained voices for both gender adults and children (p<0.05). Means of coefficient of sound differ significantly only for adults. VRP parameters and coefficient of sound were more informative, than speaking voice or aerodynamic parameters to show voice training effect. Limiting values with maximum sensitivity for trained voices of significant voice parameters were established using ROC (receiver operating characteristic) curve. Logistic stepwise regression revealed that combination of some VRP parameters with slope of speaking curve could discriminate healthy trained voice from untrained in overall 81.4% accuracy for males and in overall 87.8% accuracy for females. Similar prediction accuracy - 86.6% could be achieved in children using three VRP parameters. Formulas of probabilities for trained voice were composed.