Estimation of the glottal pulseform with inverse filtering is studied in this paper. Automatic computation of the glottal source is usually performed by applying LPC-analysis in order to estimate the vocal tract transfer function. However, the performance of linear prediction decreases as the fundamental frequency of speech increases. Therefore, a new algorithm for computing the vocal tract model is presented. The new method applies the discrete version of the Itakura-Saito distortion measure when the poles of the vocal tract filter are determined. The preliminary results when synthetic acid natural vowels were analysed are discussed shortly.