UPC ETSETB 2018 Digital Speech and Audio Processing project.
Implementation of a cover detector based on chroma features and melodies.
Evaluation of the system is carried on with covers80 dataset and other songs added by the team. D. P. W. Ellis (2007). The "covers80" cover song data set Web resource, available: http://labrosa.ee.columbia.edu/projects/coversongs/covers80/.
The structure of the song can be described as a chord progression, the most similar features we can automatically from the song are the chromas.
Implementation based on: https://librosa.github.io/librosa_gallery/auto_examples/plot_chroma.html
In order to extract the melodies automatically we will use an implementetion of Melody Extraction from Polyphonic Music Signals using Pitch Contour Characteristics.
Paper: http://mtg.upf.edu/node/2436 Implementation: https://www.upf.edu/web/mtg/melodia
Dynamic time wrapping will be used n order to compare both melody vectors(n_samples) and chroma matrices (12*n_frames).
TODOOOOOOOOO