Speech Segmentation
The process of [see page 16, splitting] a speech signal at the boundary between phonetic segments.
We define:
| Term | Meaning |
|---|---|
| Acoustic Segment | The signal range related to a single phonetic term or gap between terms. |
| Acoustic Segment Points | The point at the boundary between/separating two acoustic segments. |
| Phonetic Segment | The signal range relating to a sound (vowel/consonant). |
For splosives which often have a:
- buildup - the previous sound dies down in preparation of the stop
- stop - the energy has been kept in place
- release - the energy is released producing the explosive sound.
we place the boundaries as dotted lines to indicate they buildup from the previous sound.
See the [see page 17, example].
Due to Coarticulation, effective speech segmentation (and end-point detection) in real speech is difficult.
A simple [see page 7, speech/non-speech detector] can be constructed using STE and ZCR.