HEREDITARY PHONETIC PARAMETERS OF THE HUMAN VOICE

Voice recognition or speaker identification has been studied from several points of view. The close similarity of the voices of consanguineous persons has suggested that there must exist some hereditary phonetic parameters. To test this hypothesis, a method for characterizing the similarity or dissimilarity of voices of siblings and identical twins in numerical terms was developed. If some parameters of human voice are genetically determined, monozygotic twins must have a higher intra-pair similarity of voice than dizygotic like-sexed twins or brothers/sisters. If agreement within monozygotic twins significantly exceeds that observed in dizygotic twins of the same gender, it may be anticipated that the condition is more under genetic than environmental control. In the present work, the voices of both members of three female 21-22-year-old pairs of monozygotic twins and three female 20-24-year-old pairs of sisters speaking about the same picture for a duration of about 2 minutes were recorded by the computer program Wave Studio. The material was analyzed by the Praat 4.2 voice analyzing program: average pitch, first three formants and formant bandwidths of nine vowels, the duration of words, vowels and alveolar fricatives, word intensity, FFT spectra of vowels, broad band and narrow band spectrograms of words, and broad band spectrograms of alveolar fricatives were analysed and numerically characterized. The results, in accordance with our expectations, show that intra-pair differences between monozygotic twins were lower than those between sisters. It may be concluded that a single parameter is not capable of complete discrimination; there is no definite answer as to which parameter is the single most characteristic one of the inheritability of human voice.