The final purpose of this research is to develop a learning support system of music genre which can estimate the genre of a music from partial information from a standard MIDI file of the music. It can also provide feedback on the features of the music. To date, a standard MIDI file of 120 music titles have been identified into 4 genres, Enka, Jazz, Hard Rock and Heavy Metal by Neural Network, trained to identify these 4 genres. For an estimation experiment, subjects were asked to estimate the genre of the music they hear, with the results compared against the estimations of the system. The results show that the system has a higher judgement rate. Next, the weight of the links between each node of the Hidden Layer and that of Input or Output is observed and compared, with the significance of the results explained by an expert. From the results, certain factors from 5 of the nodes in the Hidden Layer were able to be extracted. From this, the possibility of the development of a learning support system using this particular approach for music genre is shown.
View full abstract