A Research on a System Using Deep Learning for Inferring Piano Performance

Abstract

Achieving expressive and human-like automated piano performances has proven challenging. This study proposes a deep learning system to infer expressive nuances from musical scores, addressing the limitations of traditional rule-based approaches. By leveraging neural networks to learn the mapping between scores and expert performances, the system automates the inference process, improving accuracy while enhancing efficiency. This novel application of deep learning shows promise for advancing automated music performance and enabling more artistically expressive renditions. The insights gained could have broader implications for computer-aided musical interpretation and synthesis.