Computationally efficient determination of a polynomial discriminate function using the AIC and its systolic array implementation
Youji Iiguni, Hideaki Sakai, Hidekatsu TOKUMARU · Electronics and Communications in Japan (Part III Fundamental Electronic Science) · 1990
Abstract This paper presents a computationally efficient algorithm for selecting the terms of the polynomial discriminate function (PDF) by means of the Akaike information criterion (AIC). The Levinson‐type algorithms is introduced for a computationally efficient determination of the PDF and a simple computation of the AIC. Our method is applied to seven segment and vowel recognition to show its usefulness. In an actual application, a parallel information processing is indispensable to reduction of computation time. Thus, systolic array for computing the PDF that is suitable for VLSI implementation is shown.