C_S_D 3220 Lecture Notes - Lecture 7: Spectrogram, Speech Synthesis, Formant
Document Summary
Catching a typical or ideal production of the vowel. We"re looking for the vocal tract resonances. Frequently describing by f1, f2 & f3 only. Assumes a canonical ( perfect or ideal ) form of the vowel. Anyone who"s interested in voice recognition does. You just won"t hit the same targets in context. Tense vowels have longer durations than lax vowels. Low or open vowels (open jaw position) have longer durations than high or close vowels. Tense vowels have shorter transitions to consonants. Lax vowels have longer transitions to consonants. We are ready to id vowels when canonical or in transition to another sound. Since simple target model doesn"t account for everything. Other researchers have tried to account for individual differences in vowel production (and how we can still effectively perceive vowels correctly in different idiolects) Listeners actively construct idiosyncratic vowel patterns for each speaker (research: /i/or f3 as a starting point)
