Are there any guidelines on classifying pitch contours? I've
processed speech recordings with analysis software and have the pitch
contour graphs.
Let's say I want to classify contours into either:
-Rising-falling, final
-Rising
-Rising-falling, non-final
In many instances there may be a drop/rise in frequency between start
and finish of an utterance, but I'm unsure as to what quantitive
change in frequency is relevant and determines the contour type. For
instance, what drop in frequency makes a contour a 'Rising-falling,
final' rather than 'Rising-falling, non-final'?
One approach I've tried is to calculate the change in frequency
between start and end of contour as a percentage of the start
frequency. For example, if there is a drop of around 20%+ I classify
it as final rather than non-final.
Classifying early clauses in a sentence seems easier as as the pitch
range is large, but particularly on a final clause the pitch range is
smaller so interpretating the contour as a final or non-final is quite
difficult.
Any advice?