Issues in clinical applications of bidirectional multi-step predictive analysis of speech

The topic of the presentation is an examination of several methodological problems posed by multi-step predictive analysis of speech when applied with a view to estimating vocal dysperiodicities. Problems that are discussed are the following. First, the stability of the multistep predictive synthesis filter; second, the decrease of the quantization noise by means of multiple prediction coefficients; third, the implementation of the multi-step predictive analysis via a lattice filter; fourth, the adequacy per se of the predictive analysis paradigm for estimating vocal dysperiodicities. Results suggest that implementations of linear predictive analyses that are considered to be optimal for speech coding are sub-optimal for clinical applications and vice versa. Also, linear predictive analysis per se does not appear to be under all circumstances a paradigm adequate for analysing vocal dysperiodicities clinically. An alternative is discussed which is based on a generalized variogram of the speech signal.