Tom Lentz (slides Ivana Brasileiro) Perception and Production in L2 Acquisition Week 1: A Brief Introduction to Acoustic Phonetics Lab 1 12 november 2008.ppt Tom Lentz (slides Ivana Brasileiro)
Acoustic Phonetics Physics of the speech signal Relationship between activity in the speaker's vocal tract and the resulting sounds Contrast: articulatory phonetics
What we will do… Key concepts: frequency, formants, and acoustic cues Measuring speech sound Vowels and formants (Acoustic cues: VOT) (Problems in Language Acquisition)
Frequency Cycles of vibration per second Measured in Hertz (Hz) E.g. 100 Hz = 100 repetition per second
Figure 1: Two periodic signals with frequencies of 200Hz and 400Hz Frequency Figure 1: Two periodic signals with frequencies of 200Hz and 400Hz
Sine Waves vs. Complex Sound Waves Sine waves: sounds formed by one frequency only Complex sound waves: all sounds which are not sine waves All complex sound waves can be described on the basis of the sine waves
Sine Waves vs. Complex Sound Waves b c Figure 2: three sine waves (left) and resulting complex sound wave (right)
Question If in figure 2, (a) has frequency A; (b) has frequency B and (c) has frequency C, what is the frequency of picture (d)? Answer: A
Speech Sounds Speech sounds are always complex waves Two sources of sounds Vocal folds Oral cavity Sounds produced with the vocal folds are periodic: usually perceived as voiced Sounds produced with the oral cavity are aperiodic: usually perceived as voiceless
Speech Sounds Periodic Aperiodic
Measuring Sounds Oscillogram Spectrum Spectrogram
Oscillogram Amplitude x time
Spectrum Shows all frequencies present in the signal at a given point in time
Spectrogram Combines properties of oscillogram and spectrum Measure three dimensions: time, frequency, and amplitude
PRAAT practice Task: find vocal fold vibration frequency Record your voice: which phonemes ? Analyse: how?
Formants (1) The shape of the vocal tract increases some frequencies and decreases others The increased frequencies can be seen in the spectrum as peaks; and in spectrograms as darker spots These frequency peaks are the formants
Formants (2) Formants are commonly used to describe vowels The first 3 formants (F1, F2 and F3) are important for the vowel quality Other formants (F4 and F5) are important for the naturalness of the speech sound
Vowels Acoustic Space Vowels are usually plotted in a F1 x F2 graph, since they play a prominent role in the quality of the vowel F1 in the y-axis and F2 in the x-axis This graphic represents the acoustic space of vowels, the so called vowel triangle (see handout monday)
Articulatory correlates F1 correlates with vowel height higher F1, lower vowel F2 correlates with degree of frontness higher F2, ‘fronter’ vowel
Dutch Vowels
Acoustic Cues Refers to any piece of signal that has been found by experiment to have an effect on percept (Lieberman 1996) Examples of cues are: VOT, bursts, transitions, duration and formants
Voice Onset Time (VOT) Refers to the time of voicing in relation to the consonant articulation / release burst
Acoustic Cues No one-to-one relationship between ‘cues’ and a single percept (see handout) Voicing contrast in Dutch has been shown to have about 6 acoustic correlates Cues differ in how important they are “Cue reliance” refers to how strongly listeners rely on a specific cue to identify a contrast “Cue weighting” refers to how much each cue counts
Problems in Language Acquisition Segmentation problem Coarticulation Cue ambiguity Mapping problem
Coarticulation Information about more than one sound is often encoded in the same portion of the acoustic signal
Cue ambiguity The interpretation of acoustic cues depends on their position in the signal (handout Kager) Example: VOT in English ‘pin’ ‘spin’ ‘bin’ ‘pin’: aspirated, long lag VOT (30-35ms) ‘spin’: plain, voiceless unaspirated, short lag VOT ‘bin’: voiced, short lag VOT or prevoiced
Mapping problem How do phonetic categories relate to phonological categories? Allophonic variation Contextual variation Normalization