Sampled−data representation of a nonuniform lossless tube of continuously variable length

The sound transmission through a nonuniform tube (e.g., the human vocal tract) is digitally simulated using a tube consisting of uniform segments, the length of which is determined by the sampling frequency. Then the problem arises of how continuous changes of the tube length can be represented. It is solved by inserting a segment of variable length, which causes fractional powers of z to appear in the z−transform expressions describing the transmission behavior. These are made rational by first− and third−order expansion in (z−1 − 1). For a uniform tube, the resulting error in formant bandwidth is calculated.Subject Classification: 70.50.