Automatic Extraction of Drum Tracks from Polyphonic Music Signals

  • Authors:
  • A. Zils;F. Pachet;O. Delerue;F. Gouyon

  • Affiliations:
  • -;-;-;-

  • Venue:
  • CW '02 Proceedings of the First International Symposium on Cyber Worlds (CW'02)
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

We propose an approach for extracting automatically timeindexes of occurrences of percussive sounds in an audiosignal taken from the Popular music repertoire. The schemeis able to detect percussive sounds unknown a priori in aselective fashion. It is based on an analysis by synthesistechnique, whereby the sound searched for is graduallysynthesized from the signal itself. The possibility to extractdifferent types of percussive sounds and their occurrencesin the audio signal makes it possible to build a drum trackrepresenting the essential rhythmic component of a musicpiece. We present a systematic evaluation of theperformance of our algorithm on a database of popularmusic titles. The system performs well on most of the cases(over 75%). We analyze the reasons for failure on theremaining cases, and propose solutions for yet improvingthe algorithm. The extracted percussive sounds and drumtrack serves as a basis for search by rhythmic similarity inthe context of the European project Cuidado.