Forming sequential objects

[Stephen Smoliar <smoliar@xxxxxxxxxx>]

Stephen McAdams writes:
>
>One of the important processes that gives rise to sequential objects
>is segmentation. For a theoretical treatment of this in music,
>see Lerdahl and Jackendoff's grouping preference rules.

Notwithstanding Francesco Gimi's observation that electro-acoustic music is
probably more relevant than tonal music for this problem, I think it is worth
observing that Lerdahl and Jackendoff have problems even in their own domain.
Basically, those rules seem to work for melody, which means that they can also
accommodate accompanied melody.  However, there is far more to tonal music than
accompanied tonal melody.  If the system falls apart the first time it
encounters a Bach fugue, how strong can it be?

I think there are two basic approaches we take when we try to do segmentation
in the visual domain.  One is edge detection, for which there is a vast
literature in computer vision.  There other is determining the extent of
a region of a given texture, such as a cloud or a grassy field.  That one
is a bit trickier but is again being addressed by computer vision.  Perhaps
Donald Francis Tovey's observation that counterpoint is perceived as texture
should not be taken strictly as metaphorical.  I would think that the
interesting research question is one of whether or not existing computer
vision techniques can be adapted to audio signals.

Stephen W. Smoliar; Institute of Systems Science
National University of Singapore; Heng Mui Keng Terrace
Kent Ridge, SINGAPORE 0511
Internet:  smoliar@iss.nus.sg   FAX:  +65-473-9897