The sift algorithm for fundamental frequency estimation

Abstract-In this paper a new method for estimating PO, the fundamental frequency of voiced speech versus time, is presented. The algorithm is based upon a simplified version of a general tech-nique for fundamental frequency extraction using digital inverse filtering. It is demonstrated that the simplified inverse a t e r track-ing algorithm (hereafter referred to as the SIFT algorithm) encom-passes the desirable properties of both autocorrelation and cepstral pitch analysis techniques. In addition, the SIFT algorithm is com-posed of only a relatively small number of elementary arithmetic operations. In machine language, SIFT should run in several times real time while with special-purpose hardware it could easily be realized in real time. 1. Autocorrelation, Cepstral, and Inverse Filter Analysis One of the oldest digital methods for estimating the fundamental frequency (Po) of voiced speech is auto