Machine learningFeature extraction

Melody Extraction

Melody extraction is the task of automatically isolating the main melodic contour from polyphonic music recordings. It originated from music transcription research in the 2000s and addresses the core challenge of human pitch perception: identifying the perceptually dominant pitch when many instruments play simultaneously. Modern approaches use deep learning and are essential for music analysis, cover song detection, and music-to-lyrics alignment.

Open in MethodMindSoonVideoSoon

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Sources

  1. Salamon, J., & Gómez, E. (2014). Melody extraction from polyphonic music signals using pitch contour characteristics. IEEE Transactions on Audio, Speech, and Language Processing, 20(6), 1759-1770. DOI: 10.1109/TASL.2012.2188541
  2. Klapuri, A. (2008). Automatic music transcription as we know it today. Journal of New Music Research, 33(3), 323-337. DOI: 10.1076/jnmr.33.3.323.27169
  3. Bittner, R. M., McVicar, M., Salamon, J., & Ellis, D. P. (2017). An analysis of lead and accompaniment separation in polyphonic music. In Proceedings of the International Society for Music Information Retrieval Conference. link

Related methods

Referenced by

ScholarGateMelody Extraction (Melody Extraction Algorithm). Retrieved 2026-06-04 from https://scholargate.app/en/music-information-retrieval/melody-extraction