Transcription of bell chiming recordings

Goal

We are exploring automatic transcription of bell chiming recordins. Bell chiming is a Slovenian folk music tradition involving performers playing tunes on church bells by holding the clapper and striking the rim of a stationary bell. The tunes played consist of repeated rhythmic patterns into which various changes are included.

Our goal is to design robust automatic algorithms for transcription of bell chiming recordings that would enable search and retrieval of this part of Slovenian folk music heritage within the Ethnomuse digital archive.

We also develop algorithms for transcription of other bell music, such as bell clock melodies (see papers below).

Algorithm

Our algorithm for transcription of bell chiming recordings is based on non-negative matrix factorization. Because the sounds of bells are inharmonic and their tuning not known in advance, we perform transcription in two steps. First, by analyzing the covariance matrix of the time-frequency representation of a recording, we estimate the number of bells and their approximate spectra using prior knowledge of church bell acoustics and bell chiming performance rules. Transcription is then performed by applying a non-negative matrix factorization algorithm with selective sparsity constraints to learn the basis vectors that approximate the previously estimated bell spectra, while adapting the number of basis vectors during learning.

Transcription examples

• performance on 3 bells
  Source   Transcription
• performance on 3 bells
  Source   Transcription
• performance on 4 bells
  Source   Transcription
• performance on 4 bells
  Source   Transcription
• performance on 5 bells
  Source   Transcription
• performance on 5 bells
  Source   Transcription

References

For more details, see:

• M. Marolt, "Automatic transcription of bell chiming recordings," IEEE transactions on audio, speech, and language processing, vol. 20, iss. 3, pp. 844-853, 2012.
  [Bibtex]

  @article{8992340,
  author={Matija Marolt},
  year={2012},
  pages={844-853},
  volume={20},
  title={Automatic transcription of bell chiming recordings},
  journal={IEEE transactions on audio, speech, and language processing},
  number={3},
  }

• M. Lefeber, P. van Kranenburg, and M. Marolt, "Identificatie van repertoire van achtiende-eeuwse bellen- speelklokken op basis van automatische vergelijking," Tijdschrift van de Koninklijke Vereniging voor Nederlandse Muziekgeschiedenis, vol. 61, iss. 12, pp. 147-161, 2011.
  [Bibtex]

  @article{9286996,
  author={Marieke Lefeber and Peter van Kranenburg and Matija Marolt},
  year={2011},
  pages={147-161},
  volume={61},
  title={Identificatie van repertoire van achtiende-eeuwse bellen- speelklokken op basis van automatische vergelijking},
  journal={Tijdschrift van de Koninklijke Vereniging voor Nederlandse Muziekgeschiedenis},
  number={12},
  }

• M. Marolt, "Non-negative matrix factorization with selective sparsity constraints for transcription of bell chiming recordings," in SMC 2009 : proceedings of the 6th Sound and Music Computing Conference, 23-25 July 2009 Casa da Música, Porto - Portugal, 2009, pp. 137-142.
  [Bibtex]

  @conference{7224404,
  author={Matija Marolt},
  year={2009},
  pages={137-142},
  title={Non-negative matrix factorization with selective sparsity constraints for transcription of bell chiming recordings},
  booktitle={SMC 2009 : proceedings of the 6th Sound and Music Computing Conference, 23-25 July 2009 Casa da Música, Porto - Portugal},
  }

Bell chiming performers (image from celinka.si)

 

Example factorization from estimated clusters