Institut für Nachrichtentechnik

Univ.-Prof. Dr.-Ing. Jens-Rainer Ohm

Navigation

• 1.1: Home
• 1.2: Team
  • 1.2.1: Prof. Ohm
  • 1.2.2: Dr. Rohlfing
  • 1.2.4: Sekretariat Frau Szelag
  • 1.2.5: T. Claßen
  • 1.2.6: D. Mehlem
  • 1.2.7: C. Jansen
  • 1.2.10: I. Reißel
  • 1.2.11: G. Kaschel
  • 1.2.12: H. Flasche
  • 1.2.13: Alumni
    • 1.2.13.1: Prof. Lüke
    • 1.2.13.2: Prof. Aschoff
    • 1.2.13.3: Dr. Wien
    • 1.2.13.4: Dr. Alirezaei
    • 1.2.13.6: Wiss. Mitarbeiter
    • 1.2.13.7: Diplomanden
    • 1.2.13.8: Webseiten ehem. Mitarbeiter
      • 1.2.13.8.1: M. Asbach
      • 1.2.13.8.2: H. Bakhshi-Golestani
      • 1.2.13.8.3: J. Ballé
      • 1.2.13.8.4: J. Becker
      • 1.2.13.8.5: M. Bläser
      • 1.2.13.8.6: Ch. Bulla
      • 1.2.13.8.7: O. Chubach
      • 1.2.13.8.8: H. Crysandt
      • 1.2.13.8.9: Ch. Feldmann
      • 1.2.13.8.10: V. Gnann
      • 1.2.13.8.11: K. Hanke
      • 1.2.13.8.12: I. Heisterklaus
      • 1.2.13.8.13: C. Heithausen
      • 1.2.13.8.14: F. Jäger
      • 1.2.13.8.15: P. Hosten
      • 1.2.13.8.16: A. Jose
      • 1.2.13.8.17: S. Kamp
      • 1.2.13.8.18: X. Li
      • 1.2.13.8.19: M. Meyer
      • 1.2.13.8.20: T. Meyer
      • 1.2.13.8.21: N. Qian
      • 1.2.13.8.22: T. Rusert
      • 1.2.13.8.23: J. Sauer
      • 1.2.13.8.24: J. Schneider
      • 1.2.13.8.25: M. Spiertz
        • 1.2.13.8.25.1: Publication Websites
          • 1.2.13.8.25.1.1: Single Sensor Source Separation for Acoustical Machine Diagnostics

          • Current page is 1.2.13.8.25.1.2: Beta Divergence for Clustering

          • 1.2.13.8.25.1.3: Clustering for Monaural Blind Source Separation
      • 1.2.13.8.26: A. Stojanovic
      • 1.2.13.8.27: U. Thakur
      • 1.2.13.8.28: M. Unger
      • 1.2.13.8.29: B. Zhang
• 1.4: Lehre
• 1.5: Forschung
• 1.6: Projekte
• 1.7: Publikationen
• 1.8: Software
• 1.10: Abschlussarbeiten
• 1.12: Bibliothek
• 1.13: Sammlung
• 1.14: FRIENT e.V.
• 1.15: Sitemap
• 1.16: English Pages

Beta Divergence for Clustering in Monaural Blind Source Separation

128th AES Convention

Conference homepage

Abstract

General purpose audio blind source separation algorithms have to deal with a large dynamic range for the different sources to be separated. In our algorithm the mixture is separated into single notes. These notes are clustered to construct the melodies played by the active sources. The non-negative matrix factorization (NMF) leads to good results in clustering the notes according to spectral features. The cost function for the NMF is controlled by a parameter beta. The choice of beta depends on the dynamic difference of the sources. The novelty of this paper is to propose a simple classifier to adjust the parameter beta to current dynamic ranges for increasing the separation quality.

Paper:SpGn10.pdf

Poster: Poster_AES2010.pdf

Separation Examples

• Triangel - Vibraphone
• Speech - Piccolo
• Singing - Claves
• Guitar - Clarinet
• Bass drum - Contra bassoon
• Accordion - Flute

• All zip files contain mixtures at 7 different dynamic differences:

  DD1

  dynamic difference -18 dB

  DD2

  dynamic difference -12 dB

  ...

  ...

  DD7

  dynamic difference +18 dB

• For each dynamic differences all three clustering algorithms are marked with:

  c1

  clustering with beta=1

  c2

  clustering with beta=2

  c3

  clustering with adaptive beta

• For each clustering the separated sources are marked with s1 and s2

(C) by Martin Spiertz - 28. May 2010 - spiertz@ient.rwth-aachen.de