Genre-basierte automatische Komposition von MIDI-Daten

This master thesis dwells upon the creation of musical compositions which can be matched to specific genres. It is assumed that a set of characteristics and musical motifs can be used to obtain this goal. For this purpose different criteria of a group of MIDI data are analysed and monophonic and polyphonic motifs are extracted by the use of pattern recognition techniques. The obtained data and a genetic algorithm are used to create new compositions which are related to the original songs. The developed system uses a beat-based approach. In this thesis several musical and algorithmic concepts are introduced first. Second the analysis of the data and the concept for new compositions is presented. Finally the results received through interviews of a small group of people are shown. More than 50% of the used musical pieces were correctly assigned by the test persons. Also the quality of the compositions were positively ratet.