The paper introduces a technique for online symbolic music alignment using reinforcement learning. The reinforcement learning agent, an attention-based neural network, is trained to iteratively predict the current score position from limited score and past performance contexts. The agent processes a purely pitch-based representation, with timing information incorporated in a post-processing step. This technique allows for the alignment of MIDI performances to their corresponding MusicXML scores by matching individual notes of each version. The proposed model outperforms a state-of-the-art reference model of offline symbolic music alignment.
Publication date: 4 Jan 2024
Project Page: ?
Paper: https://arxiv.org/pdf/2401.00466
