Analysis of vocal patterns as a diagnostic tool in patients with genetic syndromes

Recently it was proved that some genetic syndromes may have a specific language phenotype. In this work we apply acoustical analysis to the discrimination between four genetic syndromes: Down, Noonan, Costello and Smith-Magenis. The analysis is performed with Praat and BioVoice tools. Several estimated acoustical features are applied as input to machine-learning models. Though preliminary, the results are encouraging: the acoustical analysis of the sustained vowel /a/ give an average accuracy > 50% with both tools. Our findings confirm that for some syndromes a specific “vocal phenotype” exists that might support the clinician in highlighting syndrome’s characteristics not yet studied.