Fighting to the beat: Multi-instrumental adaptive background music approach

Abstract

This paper presents our approach to enhance the background music (BGM) in DareFightingICE, a fighting game research platform, by adding a rule-based adaptive BGM. The adaptive BGM consists of three different groups of instruments playing the BGM provided in the DareFightingICE platform. The BGM adapts by changing the volume of the groups of instruments. Each group is connected to a different element of the game. We then run experiments to evaluate the adaptive BGM by using a deep reinforcement learning AI that only uses audio as input (Blind DL AI). The results show that the performance of Blind DL AI improves while playing with the adaptive BGM as compared to playing with a baseline non-adaptive BGM. We also conduct two user studies, a perception study, and a playing study. The results from these studies show that not only human players can recognize the adaptation of the BGM, they also perform better with adaptive BGM as compared to non-adaptive BGM. Lastly, the adaptive BGM scores higher than the non-adaptive on audio aesthetics and enjoyment factors making it overall a better BGM.