Automatic generation of pitch contour for speech synthesis

The rules for the generation of the pitch contour of a fragment of the Italian language are presented. The generative model, of which the input is a formal description of the sentence and outputs the pitch contour, operates as an automaton. The basic elements for the pitch curve generation are the segments between two accents. Inside these segments the pitch curves is approximated by means of simple primitives. The pitch rules are used to control a digital speech synthesiser.