This project studies the variability of articulatory speech movements. More specifically, we are interested in how this variability varies with age, i.e. if children vary their speech movements more than adults. There is a substantial body of work on variability in speech production as a function of age that has used acoustic recordings and examined segment durations. Much less is known about variability of articulatory movements. Using state of the art recording technology and digital signal processing, we can extend our knowledge in several aspects.
(Schönle, P. 1988. Elektromagnetische Artikulographie. Berlin: Springer; Perkell, J., Cohen, M., Svirsky, M., Matthies, M., Garabieta, I., & Jackson, M. 1992. Electromagnetic midsagittal articulometer (EMMA) systems for transducing speech articulatory movements. J. Acoust. Soc. Am. 92: 3078-3096).
This method is based on the principle that when a coil moves inside a magnetic field, a voltage is induced in the coil, which is proportional to the distance between the coil and the transmitter coil generating the magnetic field. The articulograph AG500 in the Humanities Lab enables registration of speech movements in three dimensions. The registered signals are very well suited for digital signal processing.
- Functional Data Analysis (FDA)
(Ramsay JO, Silverman BW. Functional Data Analysis. 2. Springer-Verlag; New York: 2005; Ramsay et al. J. Acoust. Soc. Am. 99: 3718–3727, 1996)
This technique allows analysis of variability in time and amplitude of articulatory movements during a short phrase. This is in contrast to another technique, the Spatiotemporal Index (STI, Smith et al. Exp. Brain Res. 104; 493-501, 1995) that only provides a single, combined measure.
- Do temporal and amplitude variability decrease at the same rate with age?
- Are there differences in movement variability between children with typical and atypical language development?
- Is movement variability related to cerebellar function, assessed by the blink reflex?
Building on our experiences with using FDA for speech movements, we are further exploring FDA as a method for speech analysis and resynthesis. FDA makes it possible to reduce the statistical analysis of a set of speech parameter contours in time, like f0, formants, intensity – on their own or together. By combining FDA with other analysis methods and applying them to speech and also human–animal communication, we hope to learn more about the temporal aspects of speech and communication.