9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

DC-Constrained Linear Prediction for Glottal Inverse Filtering

Paavo Alku, Carlo Magi, Tom Bäckström

Helsinki University of Technology, Finland

Closed phase covariance (CP) analysis is a glottal inverse filtering method which estimates the vocal tract during the glottal closed phase. Since closed phase durations are typically short, the vocal tract computation with linear prediction is vulnerable to the covariance frame position. This study proposes a modified CP algorithm in which a DC-gain constraint is imposed in optimizing the linear predictive inverse model of the vocal tract. Results show that the algorithm improves the robustness of the CP analysis on the covariance frame position.

Full Paper

Bibliographic reference.  Alku, Paavo / Magi, Carlo / Bäckström, Tom (2008): "DC-constrained linear prediction for glottal inverse filtering", In INTERSPEECH-2008, 2861-2864.