View article

To say that acoustic measurements on high-pitched voices involve many difficulties, is asking the obvious. Nevertheless, the need of acoustic data on infant voices is growing as the interest in speech developmental processes increases. Analysis of the acoustic signal is a necessary means to determine the variability and similarities within normal speech development as well as in disorders. The availability of a number of audiorecordings of a hearing-impaired infant coincided with our need to test and select reliable analysis systems, suitable for the acoustic analysis of the sound productions of infants and young children. In this pilot study we concentrated on the analyses of formant frequencies of the vocalic parts of the infant's sound productions. Approximately 300 non-cry vocalisations of a hearing-impaired boy at the age of9, 12, and 15 months were analysed spectrographically, and by means of a conventional as well as a robust method of linear prediction coding. In comparing our results based on these three techniques mutually and with data from literature we have to conclude that the robust linear prediction method yields extremely high second formant frequency values, whereas the other two techniques agree within a relatively small frequency range.

As for the achieved formant frequency data of the infant we feel that at this stage it is not clear yet which aspects are the result of the analysing techniques under test and which of the vocal development of the infant. Nevertheless, a satisfying agreement is found with data from literature.