Next: Labelling / annotation
Up: Post-recording phase: transcriptionsegmentation,
Previous: Transcription
Segmentation is the process of marking the boundaries between phonemic,
phonetic or other temporal segments of a speech signal.
The process of segmentation always involves idealisations and simplifications,
because the speech signal generally only has a very indirect relation to perceived segments.
The following factors complicate the relation:
- The continuousness of most transitions between segments in the speech signal.
- The co-articulation of different vocal organs with different and partly independent timing patterns.
- The reduction (shortening, neutralisation of qualitative differences, disappearance) of segments relative to some canonical norm, particularly in fast speech.
- The assimilation of properties of a sound to the related properties of a neighbouring sound.
Segmentation is performed with signal editing software.
Dafydd Gibbon, Mon Dec 21 10:23:16 CET 1998