Page  345 ï~~PROBLEMS TO BE FACED BI BASED AUTOMATIC Alastair Clarke & Malcolm Brown Department of Computing Mathematics University of Wales College of Cardiff Mathematics Institute, Senghennydd Road Cardiff CF2 4AG Wales, United Kingdom. ABSTRACT: Several attempts have b, automatic music recogniser over the last music printing or compiling a musical dat mercially available that would help entr describes problems that have been ignored encountered during the development of a solutions to these problems will have recognition scheme for music wvi11 becom

Page  346 ï~~Other problcms Another potential problem is the recognition of major consideration with many systems only be typcstyIcs. This is less of a problem in music rec tended to be more general in nature, looking for I to differ between different musical fonts. Howe' older printed music, where symbols do not alway! modern printed edition. For example, in the exar the quaver tails, crotchet rests and bass clefs. -w----- ~J. ' a ~?-v4e.- a Az:,.~ 'F V * s~ ~ ~ j 'A -0 r,..I Top Musical example from old printed edition. 4 It should be mentioned at this point that none I..,,.,. I..--:.,..,....-..,,....-.l......,. -. I..

Page  347 ï~~the stave lines and classifying the symbols differ whilst others consider the processing time rcquircd. I UV aI Musical Example before and after stave line rem The literature describing these various existir success rates of 90-100% in recognising the sym "correctly recognised" all the musical symbols, and. rate of "nearly 100%".. However, these results werc idealised images (four in the case of Prcrau, and ten we shall describe below were not encountered. Inc solutions these problems perhaps the rea commercially available. Non-searated Svmbols One of the most common problems that we have en where two or more musical symbols have been t