Naonari Hoshide
Instrumental artists have considered how to communicate their feelings in playing the instrument. Environment accurately indicating what listeners would feel to listen an instrumental performance hastens development of artist's skill for representation. The current study argues that the system which carries the supervised learning function on which systematizes the correlation between sound features and human feelings with feedbacks from listeners can predict instrumental performance impressionis which listeners would feel. Achievement of the accurate prediction requires examination of adaptability of both a set of sound feature variables and learning algorithm. The system can help instrumental artists build a feedback loop for experiential representation practice.