Authors | Georgi Dzhambazov, Andrés Ferraro |
Contacts | {georgi.dzhambazov,andres.ferraro}@upf.edu |
Affiliation | Music Technology Group |
Hack demo | http://dunya.compmusic.upf.edu/hamr-hack/ |
Code | Github Link |
A tool which lets the user enter a melodic query by specifying a sequence of musical intervals.
It automatically finds all occurrences of the melodic query within a the vocal line of a song. For example it can find that a quint up, followed by a minor third down, followed by a octave appears at the beginning of the chorus and end of verse.
Users who are learning singing might find this tool useful to exercise music intervals for their favorite songs.
based on algorithms:
[1] Melody extraction
Note segmentation is performed based on pitch contour characteristics (island building) and signal RMS. Notes below an adjustable minimum duration are rejected. [2]
1. Query a sequence of intervals by selecting them in a grid where each step is a semitone.
2. Pitch detection and note segmentation is run behind the scenes
3. The retrieved audio segment is played and its segmented contour displayed
[1] Salomon et al. Melody Extraction from Polyphonic Music Signals using Pitch Contour Characteristics http://mtg.upf.edu/node/2436
[2] R. J. McNab et al., “Signal processing for melody transcription,” in Proc. Proc. 19th Australasian Computer Science Conf., 1996