Nonlinear inverse filtering technique for estimating the glottal-area waveform

Publisher Website
Google Scholar
Abstract
Linear time-invariant inverse filtering of the pressure or airflow signal at the human mouth during voicing is used to provide a reasonable representation of the airflow waveform at the glottis. If the glottal impedance, as it varies during the glottal cycle, does not remain much larger than the impedance of the supraglottal vocal tract, the resulting glottal-supraglottal acoustic interaction may make the glottal airflow a poor representation of the motion of the vocal folds. An inverse filter was described for deriving from oral airflow the Norton equivalent airflow of the glottal source, which was a better estimate of the glottal area than the actual airflow. This inverse filter was nonlinear in that certain filter parameters were varied dynamically, so as to compensate for the time variation of the glottal resistance and inertance. Using a feedback procedure, estimates of these glottal parameters were obtained from the filter output waveform. The nonlinera glottal-supraglottal interaction may be a significant factor in determining the spectral content of the glottal pulse and for some types of voice, in causing the primary excitation of the vocal tract to occur at the closing of the vocal folds.

This publication has 2 references indexed in Scilit: