home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!munnari.oz.au!yoyo.aarnet.edu.au!sirius.ucs.adelaide.edu.au!levels!8321207d
- From: 8321207d@levels.unisa.edu.au
- Newsgroups: sci.math
- Subject: Matching algorithm
- Message-ID: <18353.2a93ce73@levels.unisa.edu.au>
- Date: 20 Aug 92 07:25:15 GMT
- Organization: University of South Australia
- Lines: 44
-
-
- Hello folks, this is a first for me.
-
- Can anyone help with the following (sorry for being verbose).
-
- Suppose you have a number of choirs with different numbers of people in
- each. The voice ranges of the choir members are known. It is possible
- that the ranges of some of the people overlap; for example, person A may
- sing in the range 2000-2500 Hz while person B sings in the range 2300-
- 2650 Hz. Furthermore, some of the people in a particular choir may have
- other links between them, for instance person C could sing one octave
- (whatever an octave is) higher than person A.
-
- Suppose now that one of the choirs performed and this performance was
- recorded. If the recording of the performance is analyzed then it is
- possible to isolated various bands within which singing took place.
-
- The problem is to identify which choir was singing. However, care must
- be exercised during this process because when a particular voice band
- gets associated with a paerson in a choir, then there are consequences
- due to possible links which may exist between the people. Hence associating
- a voice band of 2350-2450 Hz with person A may result in the singing
- range for person D being restricted to a new range if person D sings
- at say 1.5 times that of person A. This may mean that no voice band can
- subsequently be matched with person D based on the initial match of
- person A.
-
- A scoring function for the quality of the match exist and the objective
- is to maximise the score. It has been suggested to me that this falls
- in the category of being a matching problem as found in Graph Theory.
- A check through available literature had matching problems which did
- not make use of additional constraining information.
-
- My query is whether a matching algorithm which makes use of side
- constraints exists and if so what are some pointers to it?
- Also could this be solved (or a valiant attempt made at doing so)
- by some other scheme aside from exhaustively considering every possible
- pairing between person and voice band?
-
- Any information gratefully accepted.
- Contact via email or post
-
- Cheers
- Paul
-