- Leonardo Lancia
Speech unfolds in time and there are many reasons to move away from a static characterization of speech patterns. This is a technical issue with deep theoretical implications, because static characterizations are not informative enough if we want to study speech in a dynamical framework. To model speeech data as multivariate trajectories, we followed Morriss and Carroll (2006) and used functional mixed models based on isomorphic transformations of the oberved data. Curves or images were represented by configurations of independent coefficients. These coefficients were the dependent variables in multivariate Bayesian mixed models with fixed and random factors.
Lancia, L. & Tiede M. (2012). A survey of methods for the analysis of the temporal evolution of speech articulator trajectories, in Fuchs, Weihrich, Pape, Perrier (ed.s) Speech Planning and Dynamics. (Peter Lang).
Lancia, L., Fuchs, S., & Rochet-Capellan, A. Quantification of speech convergence through non-linear methods for the analysis of time-series. International Symposium on imitation and convergence in speech, Aix en Provence, France, September 2012 (Poster presentation)