Music chord recommendation of self composed melodic lines for making instrumental sound

摘要

Music chords are important to enable musical instruments to be played harmonically. Therefore, appropriate chord determination is essential. However, automatic chord determination using only monotonic melodic information is very difficult because chords can be subjectively variable. In this paper, we propose a method for the appropriate basic chord recommendation of melodic lines. The proposed method functions by using the notes feature of each bar based on the theory of harmony. The feature is defined by the pitch and duration of each note. Also, two chords in a bar can be estimated by using bar segmentation policy based on statistical analysis. The experimental results showed that compared with the original songs, an overall accuracy of 92.04 % (81 recommended chords/88 bars) was achieved. Subsequently, the proposed method was applied to our self-composed songs for which chords had not been determined. The results confirmed that the chords derived by our proposed method were fairly harmonious with the songs.