Phil Taylor rebutted some of John Walsh's arguments with:
> Consider the problem of repeats; for a tune with a repeated first
> part, one transcriber will simply enter the part and stick a repeat
> sign at the end, ignoring the fact that the number of beats in the
> last bar does not match the anacrusis. A second, more careful
> transcriber will put in first and second endings so it plays properly.
> A third transcriber leaves the repeat sign out (the first part of
> a tune of this genre _always_ repeats, and we musicians don't need
> to be told that). A fourth transcriber is making an exact transcription
> of an actual performance, and writes the repeat out in full in order
> to notate some subtle variations of ornamentation. No method which
> depends on recognising the overall shape of the tune can see these
> four versions as being essentially the same.
>
I think that's why I began by suggesting applying it to fragments rather
than the whole tune. I haven't yet flogged through John's long
post with the Gramm Schmidt stuff (my maths IS up to it) but
I think he might claim that WITH THE RIGHT BASIS FUNCTIONS
the similarity will still drop out. If the "signal" is in the 4 bar phrases
and the 8 bar structure which is notated differently is noise then
the method should spot the similarity in the higher frequency
components. I am deeply sceptical.
To subscribe/unsubscribe, point your browser to: http://www.tullochgorm.com/lists.html
