A Priori Digital Speech Analysis
S. Boll · Defense Technical Information Center (DTIC) · 1973
The paper describes a priori digital speech analysis. This method takes advantage of the quasi-periodic nature of voiced speech sounds to reduce the amount of computation needed to obtain the analysis parameters. The paper is divided into five main areas: A digital model of the human speech apparatus and its analysis by linear predictive coding is described. The specific details of an analysis-synthesis system which was developed to test the a priori method are presented. The description of how the predictor coefficients can be estimated using a priori least square is given. Pitch detection using a priori methods is described and compared with two other popular methods. The methods of classical, a priori, and recursive least-squares estimation are developed, both from an algebraic and geometric viewpoint. (Modified author abstract)