A new definition of the distortion matrix for an audio-to-score alignment system

This work addresses audio-to-score alignment for music processing applications, but appears incremental as it builds on existing DTW-based frameworks with a modified distortion matrix.

The authors tackled the problem of audio-to-score alignment by proposing a new distortion matrix definition for a DTW-based score following framework, which involves arranging score information into note combinations and learning spectral patterns for each using instrument models, then computing the distortion matrix with these patterns and a novel signal decomposition.