Pattern-Based MIDI Composing

This master thesis deals with different approaches for algorithmic composing. Such techniques are merged with methods of pattern recognition. The aim of the thesis is the generation of new pieces of music, which are derived from existing musical content. For this purpose short and concise note patterns, called motifs, are extracted from arbitrary musical input. These patterns are used as the basis for approaches of algorithmic composing. The generated musical pieces should sound as harmonic as possible. As part of this master thesis, a software-system was developed, which can perform the task of identifying and extracting motifs in music. Furthermore, two contrary techniques of algorithmic composing, which both use the extracted motifs as input, were implemented in the software. Based on those composers, new pieces of music can be generated. Those compositions employ the MIDI standard. Furthermore, as an introduction to the world of motif extraction and algorithmic composing, different approaches and techniques are explained in detail in the literature survey.
Subsequently, the design of the implemented software is presented.
Eventually, the results are evaluated, based on a quantitative survey.