Automatic Reduction of Vowel Spectra: An Analysis‐by‐Synthesis Method and Its Evaluation

An analysis‐by‐synthesis procedure for the automatic reduction of vowel spectra is described. The scheme is realized as a program for a general‐purpose digital computer. The output of the program specifies time samples of vocalic utterances in terms of parameters that characterize the transfer function of the vocal tract. The performance is described and evaluated in terms of deviations of formant‐frequency values and magnitudes of error scores; manually derived data provided by a related scheme are used as criteria for comparison.