Phonetics: the generation of speech Phonemes “The shortest segment of speech that, if changed, would change the meaning of a word.” hog fog log *Phonemes are not letters*
The Phonemes of North American English
Speech Production: The Vocal Tract
The vocal cords generate many frequencies. Resonance with the vocal tract emphasizes certain frequencies. To make different vowel sounds, you change the shape of your vocal tract (Also try it with a whisper)
Different vowels are “located” at different positions.
Consonants are characterized by place, manner, and voicing Place - where airstream is blocked or restricted bilabials milk labiodental fast dental thing alveolar top palatal chew velar good
Manner - how air is constricted stops put / but fricatives fence /stop affricates check / jump nasals mind / note liquids rent / lent glides white / you (semivowels) Voiced/unvoiced: Bart / part; dead / Ted
Depicting sounds Basic waveform Time Amplitude
Speech spectrogram. Darkness indicates intensity of sound at a given moment at a given frequency. Silence before the “k” sound The “s” sound is broadband noise Vowels contain formants due to resonances in vocal tract.
Different vowels have different formant positions
Formant transitions characterize consonants; formant positions characterize vowels. Formant transitions Formants
Some formant transitions that occur over different lengths of time
Coarticulation: successive phonemes interact
In normal speech we run words together. The speech signal is hard to segment. Idealized word-by-word speechReal speech. Note: even when you speak clearly and correctly, the sounds run together. A tough problem for speech recognition with “connected speech.”
Categorical perception. Discrimination is best at a category boundary.
Categorical perception: sounds like“da” sounds like“ta” Gradual changes in physical stimulus lead to abrupt changes in percept. Test by synthesizing speech from the spectrogram, and adjusting the physical parameters. There is a category boundary between “da” and “ta.”
Do infants have categorical perception of phonemes? Eimas measured sucking rate to see if babies were more interested when sounds crossed category boundaries. Infant studies by Eimas. Conclusions remain controversial.
McGurk effect: vision & speech interact.
Specialized areas in the brain Left hemisphere is crucial in speech. Damage to Broca’s area leads to difficulty in speech production. Damage to Wernicke’s area leads to difficulties with understanding and to production of “word salad.”