Rudi Villing, Tomas Ward


The P-centre (or Perceptual Attack Time) is hypothesised to be:
The perceptual "moment of occurrence" of a sound

That which is regular in a perceptually regular sequence of sounds

P-centres and rhythm (by definition)
Rhythm of a sequence of sounds is given by the interval between P-centres
Applies to perception and production

Typical P-centre assumptions

The P-centre is a single unique location (not a region) in a sound
The p-centre is context independent (e.g. doesn't depend on neighbouring sounds in a sequence)

All the models being reviewed make these assumptions (though Pompino-Marschall does suggest two conflicting features exist)

Many unresolved questions (not addressed in this work)
E.g. Are P-centres a feature of all sounds? Is speech special?

Good P-centre model(s) required:
To accurately analyse rhythm in the natural performance of music or production of speech
To accurately construct/edit/synthesise speech or music with a specific rhythm

Several existing models
Are their predictions similar or different?
Do their predictions match subjective perception?

No published comparison or evaluation of all models exists (to our knowledge)
Most recent model published in 1997
This work is in progress