A new glottal wave analysis method, Pitch Synchronous Iterative Adaptive Inverse Filtering (PSIAIF), is presented. The algorithm takes advantage of a previously developed glottal wave analysis technique, Iterative Adaptive Inverse Filtering (IAIF). The basic idea in the IAIF-algorithm is that the average effect of the glottal excitation to the speech spectrum is first estimated using an adaptive low-order all-pole filter. The estimate for the vocal tract is then computed by applying LPC-analysis to the signal from which the estimated glottal contribution was eliminated. The glottal excitation is obtained by eliminating the effects of the vocal tract and lip radiation by inverse filtering. The new PSIAIF-method applies first the IAIF-analysis described above pitch asynchronously to a speech signal. The resulting glottal wave estimate is used in order to determine positions of frames for pitch synchronous IAIF-analysis. The final estimate for the glottal pulseform is obtained by analysing the original speech signal by the IAIF-algorithm using frames that span speech samples between consecutive maximal glottal openings. Preliminary results show that the PSIAIF-technique is able to give a fairly reliable estimate for the glottal excitation.
Bibliographic reference. Alku, Paavo (1991): "Glottal wave analysis with pitch synchronous iterative adaptive inverse filtering", In EUROSPEECH-1991, 1081-1084.