Source and Filter
The standard acoustic model separates speech into two parts:
- The source is the buzz from vibrating vocal folds, rich in harmonics.
- The filter is the vocal tract, which amplifies specific frequencies.
Changing the source alters pitch. Changing the filter changes the vowel.
Reading a Spectrogram
A spectrogram plots time horizontally, frequency vertically, and intensity as darkness:
- Fundamental frequency (F0): the rate of vocal fold vibration, heard as pitch.
- Formants: vocal tract resonances, seen as dark horizontal bands.
- Voicing: shown by a voice bar along the bottom.
- Silence then a burst: marks a plosive.
- Random energy at high frequencies: marks a fricative.
Identifying Vowels
Two formants identify a vowel:
- F1 varies inversely with vowel height. High vowels have a low F1, around 300 Hz.
- F2 varies with frontness. Front vowels have a high F2.
For a neutral schwa, the values sit at roughly 500 Hz and 1,500 Hz.
Voice Onset Time
VOT is the duration between stop release and the start of voicing. In English, it distinguishes /b/ from /p/.
| Category | VOT | English example |
|---|---|---|
| Voicing lead | negative | rare in English |
| Short lag | 0 to 20 ms | /b d g/ |
| Long lag | 40 to 80 ms | /p t k/ |
Perception is categorical. Listeners hear continuous variation as Two distinct categories with a sharp boundary near 30 ms.
Common pitfall: confusing F0 with F1. F0 is the source, heard as pitch. F1 is the first filter resonance, heard as vowel height.