This method is based on the acoustic tube model of the vocal tract in which the tube is divided into a finite number of sections with equal lengths. The volume velocity is defined in each section. Input impedance at the front end of the tube (corresponding to the lips) can be computed as the ratio of the pressure and the volume velocity there, providing that the volume velocity and the pressure are continuous at each boundary between two adjacent sections. This input impedance is shown to be of the formD(z)/G(z), whereD(z) andG(z) are polynomials inz−1. It is also shown thatD(z) andG(z) are constructed as a difference and a sum, respectively, of two polynomials whose coefficients can be determined from the reflection coefficients of the acoustic tube model. Since a method for obtaining the reflection coefficients from the speech wave is already established [H. Wakita, SCRL Monogr. No. 9 (July 1972)],D(z) and G(z) can thus be obtained from the speech wave. Input impedance singularities obtained with area functions for five Russian vowels by Fant were in very good agreement with Mermelstein's results [J. Acoust. Soc. Amer.41, 1283–1294 (1967)]. [The Office of Naval Research supported this work.]