The drums is one of popular musical instruments and play a fundamental role in popular music. The rhythm patterns played on drums contribute greatly to determining the features of popular music in terms of rhythm. It is assumed here that drum patterns can be distinguished into "basic-rhythm" and "fill-in" patterns. The basic-rhythm pattern is a constant pattern, similar to several repeated patterns. In contrast, the fill-in pattern is an improvised pattern compared to the basic-rhythm pattern, and is usually used to fill in blanks in the accompanying melody. No criteria for objectively detecting these patterns have not been reported yet. Therefore, this study describes a detection method based on similarity in neighborhood. The method deals with symbolic representations, as provided by MIDI data. Drum pattern is modeled as a combination of three subgroups, such as "basic group" comprised of bass drum and snare drum, "tom group", and "cymbal group". Characteristics of drums is roughly obtained by modeling it as a combination of three subgroups. The performance of the proposed method is evaluated by the experimental evaluation, and improvements of proposed method are discussed. In addition, we implemented our method as an application system that plays back MIDI data as well as detecting fill-in patterns. We will consider the features of music excerpts based on different musical genres. A drum-rhythm database was constructed during this study, and using the collected data we will work on developing an automatic arrangement system of drumrhythm patterns for given MIDI files.
展开▼