Composition of Musical Piece Suited to Natural Sound by Interactive GA using User’s EEG as Fitness

Abstract

Everyone needs to be relaxed in some way. Previous studies show that listening to the combination of musical pieces and natural sounds effectively provide listeners with a relaxed feeling. In this study, we advocate an Interactive Genetic Algorithm (IGA) that composes musical piece suited to natural sound. IGA is a method to create media content that fits users’ feelings based on their evaluation. Considering the fatigue and burden brought by subjective evaluation of users, the ratio of alpha-wave power in Electroencephalography (EEG) is used as fitness in this study. Based on the proposed IGA, we built a system taking musical notes, entire volume, and tempo as variables to adjust. With the IGA system, a listening experiment was carried out to investigate the fundamental efficiency of the proposed IGA. As a result, a marginal increase in the mean fitness was observed between the initial and final generations. However, no significant difference was observed in the subjective evaluation step after the search step.