Voiced speech synthesis with a nonlinear glottal model
Avinash Kumar, A. Gersho · 2002
We propose a model for voiced speech synthesis using a threshold autoregressive (TAR) glottal flow model. It reconstructs the pitch periodicity and tracks small pitch period variations even at a low parameter update rate of 20 ms, without requiring explicit pitch information. The model overcomes several limitations of traditional glottal models and has potential for application to high quality speech synthesis and low bit rate speech coding.