Skip to content

Code for <A Cross-Scape Plot Representation for Visualizing Symbolic Melodic Similarity, ISMIR 2019>

License

Notifications You must be signed in to change notification settings

saebyulpark/cross_scape_plot

Repository files navigation

Cross-Scapeplot Representation

Matlab implementation of "A Cross-Scape Plot Representation for Visualizing Symbolic Melodic Similarity, ISMIR 2019".

alt tag

The cross-scape plot is computed by stacking up a minimum local distance between segments from each of the two songs. After segmenting the songs, the local similarity is performed by a sequence-based similarity algorithm for all possible segments of the two songs. As the layer goes up, the segment size increases and it computes progressively more long-term distances. It is described by a hierarchical visual representation with a triangular or trapezoidal shape.

Usage

  1. Load midi files to be compared. (by MIDI toolbox 1.1)
song_a = readmidi(midi_file_path1)
song_b = readmidi(midi_file_path2)
  1. Caculate distance matrix between two songs.
dist_mat = get_distance_matrix(song_a, song_b, pitch_or_rhythm) % picth = 1, rhythm = 2
  1. Caculate similarity score from distance matrix.
score = get_similarity_score(dist_mat)

Result

result

An example of a cross-scape plot. The local similarity is indicated by color (pitch: red, rhythm: green). The x-axis shows the location of the longer melody. The y-axis represents the segment size.

References

This sample case is one of the plagiarism cases (Mood Music v. De Wolfe) and the following links provide resources for this case.

Python Implementation

Note: The Python version of this implementation has been updated, and the link to the updated code can be found here: Cross-Scapeplot_Demo.ipynb.

About

Code for <A Cross-Scape Plot Representation for Visualizing Symbolic Melodic Similarity, ISMIR 2019>

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages