Arma modeling applied to linear prediction of speech

Autoregressive Moving Average (ARMA) models present two problems in practice: (1) choice of the orders p and q of the AR and MA components respectively, and (2) estimation of the coefficients. Recently a new method of order selection or modeling of ARMA processes has been introduced by H. L. Gray and others in a yet unpublished paper "A New Approach to ARMA Modeling". In this paper we apply Gray\´s ARMA order selection method to a segment of voiced speech and compare the results to the models suggested by Akaike\´s Cannonical Correlation Analysis (CCA) and Information Criterion (AIC).