Patent US 5,522,013

A speaker recognition method, including comparing a model calculated on the basis of samples derived from a speech signal with a stored model of at least one known speaker. The averages of the cross-sectional areas or other cross-sectional dimensions of portions (C1-C8) of a lossless tube model of the speaker's vocal tract, calculated by means of the samples derived from the speech signal, are compared with the corresponding averages of the portions of the stored vocal tract model of at least one known speaker.