In popular music, the drums and the bass guitar have important roles in terms of rhythm. To investigate the role of drums, a drum pattern database has been developed employing thousands of music excerpts. However, no methods have been reported for identifying the bass guitar part from symbolic music representation. This study proposes a method to identify the bass guitar part from MIDI excerpts comprised of several instrumental parts. The proposed method identifies a bass guitar part by a heuristic approach given by several bass guitarists. Each one-bar length is identified as either a bass guitar part or not and named the bass guitar pattern (BGP). Extracts from the BGP concern onset, interval, and dynamics profiles. The onset profile represents the sequence of onset time, interval profile represents the interval for each note from root note of the excerpt, and dynamics profile represents the value rank of MIDI velocity for each note. Therefore, the BGP database constructed here consists of these three types of profile. This database is applied to analyze the above features of the bass guitar pattern. This study investigates the occurrence frequency of the BGP for each feature using the three types of profiles, along with a drum pattern database developed in another study. A practical application of this database is to automatically arrange the bass guitar part by observing the occurrence frequency for each pattern. The performance of the bass guitar part automatically arranged using the proposed method was confirmed to sound not artificial but natural.
展开▼