Piano music transcription

Transcription of music could be defined as an act of listening to a piece of music and writing down music notation for the piece. Transcription of polyphonic music (polyphonic pitch recognition) converts an acoustical waveform into a parametric representation, where notes, their pitches, starting times and durations are extracted from the signal.

SONIC is our system for transcription of polyphonic piano music. The system takes an acoustical waveform of a piano recording as its input. The output of the system is a MIDI file containing the transcription. Notes, their starting times, durations and loudness’ are extracted from the signal.

At MIREX 2016 evaluation, SONIC achieved solid results on the Multiple Fundamental Frequency Estimation & Tracking task for both the MIREX and the Su datasets..

Dowload SONIC

SONIC is available for download. This version was also submitted to MIREX 2016. Try it yourself!

Download

Implementation details

SONIC is based on a connectionist paradigm; neural networks are used for partial tracking, note recognition, onset detection and detection of repeated notes.

For more details, see the following publications:

  • [PDF] M. Marolt, "A connectionist approach to automatic transcription of polyphonic piano music," IEEE transactions on multimedia, pp. 439-449, 2004.
    [Bibtex]
    @article{4203860,
    author={Matija Marolt},
    year={2004},
    pages={439-449},
    title={A connectionist approach to automatic transcription of polyphonic piano music},
    journal={IEEE transactions on multimedia},
    }
  • [PDF] M. Marolt, "Networks of adaptive oscillators for partial tracking and transcription of music recordings," Journal of new music research, vol. 33, iss. 1, pp. 49-59, 2004.
    [Bibtex]
    @article{4404820,
    author={Matija Marolt},
    year={2004},
    pages={49-59},
    volume={33},
    title={Networks of adaptive oscillators for partial tracking and transcription of music recordings},
    journal={Journal of new music research},
    number={1},
    }
Transcription examples

Synthesised recordings

Transcriptions of recordings synthesised from MIDI files with various piano samples.

Real Piano music recordings

Transcriptions of real recordings taken from CDs.

Other non-piano recordings

SONIC is designed to transcribe solo piano music. How does it fare with other recordings?

  • A.C. Jobim, Saudade do Brasil, performed by Bill Evans (piano), Eddie Gomez (bass), Fantasy OJCCD-814-2
    Source   Transcription   Resynthesis
  • J.S. Betty, C.A. Green, Just In Time, performed by Oscar Peterson (piano), Ray Brown (bass), Ed Thigpen (drums), Verve 852 769-2
    Source   Transcription   Resynthesis
  • O. Lopez, Classical Soul, perf. Oscar Lopez (guitar), Narada 61065
    Source   Transcription   Resynthesis
  • C. Atkins, Acoustic Guitar Medley, perf. Chet Atkins (guitar), ??
    Source   Transcription   Resynthesis
  • P. Desmond, Take Five, performed by Dave Brubeck (piano), Paul Desmond (alto saxophone), Eugene Wright (bass), Joe Morello (drums), Columbia/Legacy CK 65122
    Source   Transcription   Resynthesis
  • T. Monk, Thelonious, performed by Thelonious Monk (piano), Idrees Sulieman (trumpet), Danny Quebec West (alto sax), Billy Smith (tenor sax), Gene Ramey (bass), Art Blakey (drums), Blue Note CDP 7243 8 30363 2 5
    Source   Transcription   Resynthesis
  • A. Piazzolla, Invierno Porteno, performed by Astor Piazzola (bandoneon), Antonio Agri (viola), Oscaldo Manzi (piano), Cacho Tirao (el. guitar), Kicho Diaz (bass), BMG 74321 21373-2
    Source   Transcription   Resynthesis
  • L. Brown, S. Stept, C. Tobias, Comes Love, performed by Billie Holiday (vocals), Harry Edison (trumpet), Ben Webster (tenor sax), Jimmy Rowles (piano), Barney Kessel (el. g.), Red Mitchell (bass), Gus Johnson (drums), Verve 314 513 943-2
    Source   Transcription   Resynthesis

Oscillator networks are at the core of SONIC