Modeling the control mechanism for generating the rise-fall patterns in FO contours

The strategy of modeling the control mechanism for generating F0 contour of speech signal is studied in this paper. Based on some dynamic characteristics of vocal cord strain, the complex laryngeal mechanism relative to local F0 regulation is simplified to be a feasible physical model. Furthermore, a model function is deduced as the control mechanism for the generation process of local rise-fall patterns, and two kinds of basic feature patterns result with so called rise-fall commands defined by model parameters. on the logarithmic scale of F0 versus time the local characteristics of an F0 contour are approximated by the sum of these patterns generated by appropriate commands. The experimenial results in analyzing and synthesizing the F0 contours of spoken Chinese utterances indicate that the observed F0 contours can be always approximated well by the model, and a good correlation exists between some model parameters and the transition duration of local F0 rising or falling. The model lays a foundation for Chinese F0 contour synthesis by rule.