Fifth International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2007)

Florence, Italy
December 13-15, 2007

Improvement of Source-Tract Decomposition of Speech Using Analogy with LF Model for Glottal Source and Tube Model for Vocal Tract

T. Dubuisson, Thierry Dutoit

Circuit Theory and Signal Processing Lab (TCTS Lab), Faculté Polytechnique de Mons, Belgium

In this paper we propose improvements to a recent algorithm of speech decomposition into glottal source and vocal tract contributions. This algorithm is based on the Zeros of the Z-Transform (ZZT) representation and requires restrictive conditions about the analysis window. Inaccurate results of decomposition can occur if these conditions are not fulfilled. The improvement method consists in considering an analogy with the LF model for the glottal source and a tube model for the vocal tract. Results are presented for a sustained vowel /a/ in both time and spectral domain. Future developments are also proposed.
Index Terms. Zeros of the Z-Transform, glottal source, vocal tract, speech decomposition, Glottal Closure Instant

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Dubuisson, T. / Dutoit, Thierry (2007): "Improvement of source-tract decomposition of speech using analogy with LF model for glottal source and tube model for vocal tract", In MAVEBA-2007, 119-122.