An Experiment in Automatic Isolation and Identification of Vowels in Continuous Speech
Iris Kameny, Raymond Weeks
Abstract
Iris Kameny, Raymond Weeks
Abstract
Results of a vowel isolation and identification experiment provided tentative answers to some acoustic phonetic questions concerning continuous speech. The experiment indicated that (1) automatic techniques can be employed to isolate vowels in the speech signal; (2) vowels consistently reach an approximate target in terms of their first three formants; (3) it is possible to locate a “steady-state” portion of the vowel that corresponds closely to the vowel target; (4) a single spectrum extracted in the steady-state region of the vowel locates that vowel reliably with respect to the three-dimensional formant space; and (5) comparison of the vowel formants with a previously compiled, speaker-dependent table of vowel formants provides a satisfactory method of identifying the vowel.
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Results of a vowel isolation and identification experiment provided tentative answers to some acoustic phonetic questions concerning continuous speech. The experiment indicated that (1) automatic techniques can be employed to isolate vowels in the speech signal; (2) vowels consistently reach an approximate target in terms of their first three formants; (3) it is possible to locate a “steady-state” portion of the vowel that corresponds closely to the vowel target; (4) a single spectrum extracted in the steady-state region of the vowel locates that vowel reliably with respect to the three-dimensional formant space; and (5) comparison of the vowel formants with a previously compiled, speaker-dependent table of vowel formants provides a satisfactory method of identifying the vowel.
Key concepts: Formant, Vowel, Mid vowel, Speech recognition, Acoustics, Computer science, Identification (biology), Mathematics