This paper describes an algorithm for identifying the parameters of a Vocal Tract model from a record of the lip-pressure/glottal-flow impulse response. The basis of the identification algorithm is a formal inverse to the reference Vocal Tract network. The Vocal Tract model is in the form of a digital network, topologically equivalent to the normal lossless acoustic tube model, but with frequency-dependent terminations. The topological equivalence suggests an extension of the Burg [1] algorithm for identification of a lattice network production model. This extended algorithm is described. To assess the performance of the algorithm, it was tested on impulse response records generated from the six Russian vowels of Fant[2]. The results are compared with those of Strube[3].