Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Formant Tracking Using Segmental Phonemic Information

Minkyu Lee, Jan van Santen, Bernd Möbius (1), Joseph Olive

Bell Labs, Lucent Technologies, Murray Hill, NJ, USA
(1) IMS, University of Stuttgart, Stuttgart, Germany

A new formant tracking algorithm using phoneme depen-dent nominal formant values is tested. The algorithmcon-sists of three phases: (1) analysis, (2) segmentation, and (3) formant tracking. In the analysis phase, formant can-didates are obtained by solving for the roots of the lin-ear prediction polynomial. In the segmentation phase, the input text is converted into a sequence of phonemic symbols. Then the sequence is time aligned with the speech utterance. Finally, a set of formant candidates that are close to the nominal formant estimates while satisfy-ing the continuity constraints are chosen. The new algo-rithm significantly reduces the formant tracking error rate (3.62%) over a formant tracking algorithmusing only con-tinuity constraints (13.04%). We will also discuss how to further reduce the tracking error rate.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Lee, Minkyu / Santen, Jan van / Möbius, Bernd / Olive, Joseph (1999): "Formant tracking using segmental phonemic information", In EUROSPEECH'99, 2789-2792.